Timeline

Adds a section disclosing the appointment of a Chief AI Officer who leads the Agency's AI strategy alongside its AI accountable official.

• Chief AI Officer appointed, leading the Agency's AI strategic approach across the organisation and government
• Chief AI Officer described as working in tandem with the AI accountable official (AIAO) to balance innovation with accountability, trust and safety

• + Added: Chief AI Officer: not mentioned → in place

Discloses AI agents and AI-assisted triage of enterprise agreements, names a new accountable official, and now describes its AI use as current rather than exploratory.

• New use disclosed: AI supporting initial assessment and triage of enterprise agreements, identifying and categorising matters for consideration by Commission staff
• New use disclosed: agents to automate administrative tasks, integrate information and systems, and streamline workflows
• Accountable official changed from Executive Director of Enabling Services to Executive Director, Legal Performance and Engagement
• Wording shifted from 'we are exploring the use of AI' to 'we use AI', with usage patterns (workplace productivity, analytics for insights) and domains (corporate and enabling, service delivery) now stated against the Standard's classification framework

Changes the statement's "Updated" date from 3 December 2025 to 2 October 2026, with no other change to the text.

Changes the statement's "Updated" date from 3 December 2025 to 2 October 2026, with no other change to the text.

No change to the statement; APRA's standard "about APRA" page-footer blurb leaked into the capture beneath the empty Footnotes heading.

Discloses trials of AI for decision-making and service delivery uses, softens its no-public-impact assurance to "currently", and adds a Chief AI Officer to its governance.

• Added: discloses proof-of-concept and trial work on use cases that would fall in the 'decision making and administrative action' pattern and 'service delivery' domain once implemented
• Removed: 'service delivery' from the list of activities the Entity says it does not use AI for
• Altered: assurance that AI is not used where the public may interact with or be significantly impacted by it now qualified as 'does not currently use'
• Added: Chief AI Officer named alongside AI Accountable Officials as guiding communication and training

• + Added: Domain added: service delivery
• + Added: Usage pattern added: decision making and administrative action
• ~ Changed: Stated last-updated date: 2026-07-03 → 2026-09-15

Rewrites the statement, naming its AI tools, disclosing AI support for decision-making and recruitment, adding a steering committee, and declaring no public-facing AI.

• Altered: 'does not use AI for automated decision-making' becomes 'does not use AI to make final decisions'; AI now supports decision-making and administrative action, with staff making the decision
• Added: usage patterns and domains under the Australian Government Classification System for AI Use, including service delivery and regulation and compliance
• Added: new uses disclosed, including administrative support for recruitment, generating visuals and imagery, workflow automation, and flagging anomalies for staff review
• Added: tools named: Microsoft 365 Copilot, Copilot Studio, GitHub Copilot, a Databricks AI trial, and GovAI Chat (limited to public and OFFICIAL information)

• − Removed: Commitment dropped (will not): AI will not be combined with automated decision making (ADM) programs
• − Removed: Commitment dropped (will): Developing an AI Strategy and Roadmap for the coming 12–18 months
• − Removed: Commitment dropped (will): Periodically review governance arrangements and statement to remain compliant with the Policy
• − Removed: Strategic position on AI: planned → not mentioned
• + Added: Domain added: compliance and fraud detection
• + Added: Domain added: service delivery
• + Added: Named tool added: Databricks
• + Added: Named tool added: GitHub Copilot
• + Added: Named tool added: GovAI Chat
• + Added: Named tool added: Microsoft Copilot Studio
• + Added: New commitment (will not): Staff must not enter personal, OFFICIAL: Sensitive, PROTECTED or higher classified information into GovAI Chat
• + Added: New commitment (will not): The department does not use AI to make final decisions
• + Added: New commitment (will): Developing measures to understand adoption, staff confidence, quality and benefits
• + Added: New commitment (will): Statement will be updated to reflect significant developments in arrangements and new use cases
• + Added: Usage pattern added: decision making and administrative action
• ~ Changed: Stated last-updated date: — → 2026-10

Reorganises the statement under new headings (usage patterns, governance, internal roles, next steps), fixes typos, and points its contact link at what looks like a test server.

• Contact link now points to nginx-test-nacc.govcms7.amazee.io, apparently a staging host, rather than nacc.gov.au
• New section headings: AI usage patterns, AI governance, Internal AI roles, Next steps
• Chief AI Officer and AI Accountable Official headings now sit under the AI Governance Committee heading
• Typo fixes: 'every twelve (12) months', 'that that', 'further adoption of further'

Rewrites its statement: names Microsoft 365 Copilot Chat, appoints a Chief AI Officer, says its AI strategy is now developed, and sets out responsible AI principles and governance mechanisms.

• Chief Data and Analytics Officer appointed as Chief AI Officer
• Names Microsoft 365 Copilot Chat as a staff productivity tool; staff must complete training before using approved AI tools
• Commits that AI tools won't make decisions or take administrative action without human intervention; staff must verify outputs
• Strategic position on AI adoption now developed (Digital and Data Strategy 2026–2029), replacing the promise to update the statement by 15 June 2026

• + Added: AI use-case register: not mentioned → in place
• + Added: Chief AI Officer: not mentioned → in place
• + Added: Named tool added: Microsoft 365 Copilot Chat
• + Added: New commitment (human oversight): AI does not determine compliance outcomes without human review
• + Added: New commitment (human oversight): Staff remain responsible for verifying AI-generated outputs and for all final decisions
• + Added: New commitment (will not): AI systems must not involve or result in unfair discrimination
• + Added: New commitment (will not): AI tools not approved to make decisions or take administrative action without human intervention
• + Added: New commitment (will): Every AI system must have an accountable owner and be assessed before use
• + Added: Safeguard added: human review of outputs
• + Added: Safeguard added: incident or concern reporting
• + Added: Safeguard added: risk assessment
• + Added: Safeguard added: use case register
• + Added: Staff training: available → mandatory
• + Added: Strategic position on AI: planned → in place
• ~ Changed: Commitment kept (will): Developing an AI strategy covering data governance, privacy and cybersecurity compliance
• ~ Changed: Commitment kept (will): Will develop and educate all staff with AI basics resources
• ~ Changed: Commitment kept (will): Will update the transparency statement to reflect the AI strategic position by 15 June 2026
• ~ Changed: Stated last-updated date: 2026-02 → —

NEMA's AI transparency statement is unchanged; this capture drops unrelated privacy, copyright, disclaimer and public interest disclosure text that the scraper had picked up from the page before.

Capture now begins at the AI transparency statement and drops the unrelated statutory reporting material that shared the page; the statement text itself is unchanged.

IGIS's AI transparency statement is unchanged; this capture simply drops the unrelated accountability-page material (financial reporting, registers, code of conduct procedures) that the previous capture had included.

IP Australia's statement text is unchanged; the capture moved from page-numbered PDF extraction to clean Markdown headings and paragraphs.

Rewrites its statement to let AI access internal ARPC data, stop naming Copilot and ChatGPT, name the COO Chief AI Officer, and add a use-case register.

• AI may now access internal ARPC information and data the user is authorised to see (previously limited to public data, with no direct access to internal systems)
• Removed: named tools Microsoft 365 Copilot Web and ChatGPT
• COO now designated Chief AI Officer as well as Accountable AI Official, and approves AI use cases
• New AI Use Case Register, with an accountable owner and a risk assessment before approval

• − Removed: Commitment dropped (will not): AI tools do not have direct access to ARPC's internal systems or information
• − Removed: Named tool dropped: ChatGPT
• − Removed: Named tool dropped: Microsoft 365 Copilot
• + Added: AI use-case register: not mentioned → in place
• + Added: Chief AI Officer: not mentioned → in place
• + Added: Commitment: no public-facing AI without a human added
• + Added: New commitment (human oversight): AI outputs reviewed by skilled employees before being relied on; AI-assisted code passes review and testing
• + Added: New commitment (will not): Do not permit AI systems to autonomously plan, make decisions or take actions on ARPC's behalf
• + Added: New commitment (will not): No AI decisions or administrative actions significantly affecting external stakeholders without human review
• + Added: Safeguard added: human review of outputs
• + Added: Safeguard added: testing or evaluation
• + Added: Safeguard added: use case register

Updates non-AI legal services text on the page to cite the Legal Services Directions 2025, and drops an older Senate Order contracts link. None of its AI content changed.

• Legal services text now cites the Legal Services Directions 2025 (s 109(1)(d)) instead of the 2005 Directions
• National security legal work may now go to the Attorney-General's Department as well as the Australian Government Solicitor
• Removed: link to the Senate Order for Entity Contracts, 1 January to 31 December 2025
• The captured page appears to be IGIS's corporate reporting page rather than its AI transparency statement

A stray "Search" heading from the page's navigation dropped out of the capture after the contact details; the statement text is unchanged.

Rewrites its statement. It adds analytics use and a list of approved AI uses, and describes its own governance, where it previously said the Health department managed governance.

• New use disclosed: analytics for insights in the corporate and enabling domain, as well as workplace productivity
• Four areas of approved AI tool use listed: content and communication, administration and collaboration, information management and research, and data analysis
• New commitment that AI does not make decisions or take administrative action on OTA's behalf
• New staff responsibility to check AI output before it is relied on or published

• − Removed: Commitment dropped (will): Be transparent as we explore, adopt and integrate AI technology
• + Added: AI use-case register: not mentioned → in place
• + Added: Domain added: corporate and enabling
• + Added: Monitoring and protection measures added
• + Added: New commitment (human oversight): Staff check AI-generated content for accuracy, appropriateness and compliance before it is relied on, shared or published
• + Added: New commitment (will not): AI does not make decisions or take administrative action on behalf of OTA
• + Added: Safeguard added: human review of outputs
• + Added: Safeguard added: incident or concern reporting
• + Added: Safeguard added: privacy or security controls
• + Added: Safeguard added: risk assessment
• + Added: Safeguard added: staff training
• + Added: Safeguard added: use case register
• + Added: Staff training: not mentioned → available
• + Added: Usage pattern added: analytics for insights
• ~ Changed: Policy version referenced: v1 → unspecified

DSS now discloses an AI-enabled OCR tool that converts National Redress Scheme applications into text, and says staff will check the output before any decision is made.

• New AI system disclosed: AI-enabled optical character recognition (OCR) that converts scanned and handwritten National Redress Scheme applications into machine-readable text
• New safeguard: authorised staff check the converted text against the original application before any decision or administrative action, and the original document stays authoritative
• Says the OCR tool does not make decisions or determine outcomes
• Wording on drafting policy documents softened to 'assisting with drafting'

• + Added: New commitment (human oversight): Authorised staff will verify OCR-converted Redress applications against originals before any decision or action
• ~ Changed: Stated last-updated date: 2026-07-22 → 2026-09-09

ASIC gives its accountable official and chief AI officer roles to two different executives, where one leader used to hold both, and logs the change in its revision history.

• Accountable official is now the Executive Director, Digital, Data and Technology
• Chief AI Officer is now the Executive Director, Markets, split from the accountable official role
• Removed: Senior Executive Leader, Data, Analytics and AI as holder of both roles
• Revision history entry added dated 4 September 2026

• ~ Changed: Stated last-updated date: 2026-02-27 → 2026-09-04

Moves its ML and AI principles page to a new website platform, fixing a formatting glitch and updating links, with no change to what the statement says.

PM&C says it has appointed an Accountable Official and a Chief AI Officer, names Justin Keefe as Accountable Official, and dates the statement 1 May 2026, alongside minor wording fixes.

• Discloses the appointment of an Accountable Official and a Chief AI Officer to oversee AI adoption
• Justin Keefe now identified as PM&C Accountable Official
• Statement date changed from 28 February 2025 to 'Updated 1 May 2026'
• Policy link updated and agency name corrected to Digital Transformation Agency

• + Added: Accountable official: not mentioned → designated
• + Added: Chief AI Officer: not mentioned → in place
• ~ Changed: Policy version referenced: v1 → unspecified
• ~ Changed: Stated last-updated date: 2025-02-28 → 2026-05-01

No change by the agency: the contact email that the scraper had earlier captured as a "[email protected]" obfuscation placeholder now appears in full as enquiries@auasb.gov.au.

No change by ARPC: the contact email now shows as enquiries@arpc.gov.au, where the previous capture had the "[email protected]" placeholder that Cloudflare's email protection puts in its place.

The contact address, which Cloudflare's email obfuscation had hidden in the previous capture, now shows as info@alrc.gov.au. The agency made no change.

The AI Accountable Official's contact email (aiao@aihw.gov.au) now appears in full; the previous capture had only the "[email protected]" placeholder from the site's email obfuscation.

No change by the agency: the AI contact email, previously hidden by Cloudflare's email obfuscation, now comes through in the capture as ai@aifs.gov.au.

No change by the agency: the contact email, previously hidden by the website's email-protection placeholder, now shows as enquiries@aasb.gov.au.

Discloses Microsoft Copilot use across productivity and analytics, including law enforcement and intelligence work, authorises public generative AI tools, and names its AI governance roles, including a Chief AI Officer.

• Names Microsoft Copilot as its AI tool, replacing a vague generative AI trial
• Adds the 'law enforcement, intelligence and security' domain alongside corporate and enabling
• Adds the 'analytics for insights' usage pattern, including analysis of financial records and datasets
• Lists trialled use cases: drafting documents and interview scripts, summarising legal documents, transcribing meetings, automated retention labelling

• + Added: Accountable official: not mentioned → designated
• + Added: Chief AI Officer: not mentioned → in place
• + Added: Domain added: law enforcement intelligence and security
• + Added: Named tool added: Microsoft Copilot
• + Added: New commitment (will not): The NACC will not use AI to replace human judgement
• + Added: Policy compliance statement added
• + Added: Safeguard added: governance body
• + Added: Safeguard added: privacy or security controls
• + Added: Safeguard added: risk assessment
• + Added: Safeguard added: staff training
• + Added: Staff training: not mentioned → mandatory
• + Added: Usage pattern added: analytics for insights
• ~ Changed: Stated last-updated date: 2025-10-17 → 2026-08-31

Adds a link to its Senate Order contracts report for 2025–26 on the same page; nothing in its AI transparency statement changed.

Replaces the February 2025 statement with a March 2026 update. It names the CDO as Chief AI Officer, discloses SmartGate and Microsoft 365 Copilot, adds an AI registry and adds a contact email.

• Chief Data Officer also named Chief AI Officer under the APS AI Plan
• SmartGate disclosed as the department's only public-facing AI interaction
• Microsoft 365 Copilot named as an approved generative AI tool
• New commitment: no AI system makes administrative decisions affecting individuals without human review

• + Added: AI use-case register: not mentioned → in place
• + Added: Chief AI Officer: not mentioned → in place
• + Added: Domain added: policy and legal
• + Added: Named tool added: Microsoft 365 Copilot
• + Added: New commitment (will not): Generative AI tools do not make administrative decisions about individuals
• + Added: New commitment (will): Improve AI and data skills through comprehensive training; all staff must complete AI in Government Fundamentals
• + Added: New commitment (will): Update the AI registry regularly with each system's risks and controls
• + Added: Public contact added
• + Added: Safeguard added: risk assessment
• + Added: Safeguard added: use case register
• + Added: Staff training: available → mandatory
• + Added: Strategic position on AI: not mentioned → planned
• + Added: Usage pattern added: workplace productivity
• ~ Changed: Stated first-published date: — → 2025-02
• ~ Changed: Stated last-updated date: 2025-02 → 2026-03

Adds 2025–26 legal expenditure figures and links the Q4 2025–26 gifts and benefits register. Neither change concerns AI; the captured page looks like a general corporate transparency page.

• 2025–26 legal expenditure added: $62,374 external, $734,231 internal, $796,605 total
• Gifts and benefits register link moved from the Q3 to the Q4 2025–26 register

Fixes a spelling ("focussing" becomes "focusing") in the section on its AI use by domain; the substance is unchanged.

Rewrites its statement in full, naming a Chief AI Officer, disclosing AI analysis of Consultation Hub submissions, and softening its pledge to keep AI away from public-facing impacts.

• Chief AI Officer appointed under the APS AI Plan; the role is held by the Head of Technology and Digital Policy Division
• New disclosure: AI helps analyse public submissions received through the Consultation Hub, with notice in the privacy collection statement, consent obtained and human review of outputs
• Removed: the commitment not to use AI wherever the public directly interacts with it or feels a significant impact from it without a human agent. It is replaced by the narrower statement that the public does not currently interact with its AI and that decisions get human review
• Removed: 'compliance and fraud detection' from the domains where AI is used

• − Removed: Domain dropped: compliance and fraud detection
• − Removed: Safeguard dropped: acceptable use policy
• + Added: Chief AI Officer: not mentioned → in place
• + Added: New commitment (will): Assess new uses of AI, including risks and impacts, before they are introduced
• + Added: New commitment (will): Inform individuals and obtain consent before using AI to analyse consultation submissions
• + Added: Safeguard added: privacy or security controls
• ~ Changed: Stated last-updated date: 2026-04-07 → 2026-08-25

Rewrites its statement to report the AI Strategy and Governance Framework as adopted and add a compliance checklist, but drops the named Chief AI Officer and Accountable Official roles.

• Removed: Chief AI Officer role shared by the CIO and the Regulatory Transformation Group Manager
• Removed: naming of the CIO as the AI Accountable Official
• AI Strategy 2026–27 and AI Governance Framework now reported as published internally in March 2026, not merely planned
• New compliance table for each mandatory policy requirement, with several marked 'Yes' while noted as 'In progress'

• − Removed: Chief AI Officer: in place → not mentioned
• − Removed: Legislative compliance statement removed
• + Added: AI use-case register: not mentioned → in place
• + Added: New commitment (will): Every new and existing AI use case will undergo mandatory risk assessment using DTA's AI Impact Assessment tool
• + Added: Safeguard added: staff training
• + Added: Safeguard added: use case register
• + Added: Staff training: not mentioned → mandatory
• + Added: Usage pattern added: decision making and administrative action
• ~ Changed: Commitment kept (will): Formalise strategic approach with AI Strategy 2026–27 and AI Governance Framework in March 2026

Takes the healthdirect mobile app off its list of digital services covered by the statement.

• Removed: healthdirect mobile app from the listed digital services

Names the Chief Operating Officer as its only AI Accountable Official, dropping the Chief Data & Analytics Officer from the shared role.

• Removed: Chief Data & Analytics Officer as a co-Accountable Official
• Chief Operating Officer is now the sole AI Accountable Official

Switches a heading and table label to sentence case and tidies a date's format; the policy compliance claims are unchanged.

Removes bold formatting from headings, replaces ampersands with "and" and standardises dashes, leaving the content of the statement unchanged.

AUSTRAC now says some AI interacts with the public, used to transcribe media interviews with human review. Its pilots are presented as live use, and it claims compliance with policy v2.0.

• Removed: statement that AUSTRAC has not deployed AI that directly interacts with the public
• New disclosure: limited public-facing AI, used to transcribe interviews with media entities, subject to human review
• Enterprise generative AI described as in use rather than 'trialling'
• Creation of intelligence products moved from piloting to current workplace-productivity use

• + Added: Strategic position on AI: not mentioned → in place
• ~ Changed: Policy version referenced: v1 → v2
• ~ Changed: Public-facing AI use: none → with human review

DVA classifies its AI use by domain and usage pattern, discloses trials of other staff chatbots, dates its Chief AI Officer appointment and adds a monitoring commitment.

• New disclosure: trials of other chatbots for staff queries and information gathering
• AI use classified under the DTA scheme: service delivery, policy and legal, corporate and enabling domains; analytics and workplace productivity patterns
• Semantic Text Analysis now described as supporting presumptive liability for certain claim conditions
• Chief AI Officer appointment dated February 2026

• − Removed: Usage pattern dropped: decision making and administrative action
• + Added: Domain added: policy and legal
• ~ Changed: Stated last-updated date: 2026-04-02 → 2026-08-13

Adds the Chief Operating Officer to the officials accountable for the board's AI obligations, alongside the Managing Director and Business Services Manager.

• Chief Operating Officer added as an accountable official for AI compliance

Adds the Chief Operating Officer to the officials accountable for AI compliance, alongside the Managing Director and the Business Services Manager.

• Chief Operating Officer added as an accountable official for AI obligations

Names its Manager, Data & Digital Solutions as Chief AI Officer, replacing a promise to appoint one during 2026, and moves the last-updated date to 7 August 2026.

• Chief AI Officer designated: Manager, Data & Digital Solutions
• Removed: statement that a CAIO will be appointed during 2026
• Most recent update date changed from 12 June 2026 to 7 August 2026

• + Added: Chief AI Officer: planned → in place
• ~ Changed: Commitment kept (will): Appoint a Chief AI Officer during 2026
• ~ Changed: Stated last-updated date: 2026-06-12 → 2026-08-07

Breadcrumb navigation ("Home", "Policies") from the page now shows up above the statement. The statement text itself is unchanged.

Corrects two typos, restoring "external audit committee" and "Accountable Official"; no change of substance.

Extends its Senate Order nil-contracts statement to cover 2025–26 and rewords the list; the AI content itself is unchanged.

• Senate Order nil-contracts coverage extended from 30 June 2025 to 30 June 2026
• calendar-year coverage still ends 31 December 2025; both bullets reworded (with a grammatical error)

Moves the review date to August 2026, says AI now also helps create data assets and analysis, and drops its mention of regular reviews and audits.

• Most recent review of the statement is now August 2026 (was July 2026)
• Analytics use widened: AI also helps create data assets and analysis, not only coding
• Removed: the claim of 'regular reviews and audits' and staff training programs as risk management; now a general line about 'robust governance arrangements, processes and risk management strategies'
• AI use now also governed by the CGC's internal policies as well as the DTA policy

• ~ Changed: Stated last-updated date: 2026-07 → 2026-08

Removes the "(v2.0)" version tag from the link text naming the Policy for the responsible use of AI in government; nothing else changes.

Rewrites its statement to name ChatGPT, Microsoft Copilot 365 and GovAI, classify its AI uses, and describe a new AI team, agency AI policy and staff survey.

• Names its endorsed AI tools: ChatGPT, Microsoft Copilot 365 and GovAI
• Classifies its AI use under the DTA classification system: corporate and enabling domain, used for coding analytics, research and analysis, and workplace productivity
• Discloses an AI team that explores use cases, runs an annual staff survey on AI use and holds regular training
• Discloses an agency AI policy covering its strategic position, staff guidance and what to do in an incident

• − Removed: Commitment dropped (will): Staff must complete internal AI training as a prerequisite to using AI tools
• + Added: Domain added: corporate and enabling
• + Added: Named tool added: ChatGPT
• + Added: Named tool added: GovAI
• + Added: Named tool added: Microsoft 365 Copilot
• + Added: Safeguard added: governance body
• + Added: Safeguard added: incident or concern reporting
• + Added: Strategic position on AI: not mentioned → in place
• + Added: Usage pattern added: analytics for insights

Loosens the AEC's ban on public-facing AI, names the GovAI Chat trial, closes its Copilot trial, adds AI-in-recruitment guidelines, and changes its accountable official and governance committee.

• Removed: commitment not to use AI in any AEC interaction with the public; now only bars AI from independently giving authoritative advice or making decisions in public-facing interactions
• New use disclosed: taking part in the Department of Finance's whole-of-government GovAI Chat trial with a selected group of staff
• Microsoft 365 Copilot trial reported as completed, and the tool stays available to participating staff
• Removed: stated requirement that staff complete generative AI training before using Microsoft 365 Copilot

• − Removed: Commitment dropped (will not): Staff not to use AI as part of any AEC interaction with the public
• − Removed: Commitment dropped (will): Staff must complete internal generative AI training before using Copilot
• − Removed: Safeguard dropped: staff training
• − Removed: Staff training: mandatory → not mentioned
• + Added: Named tool added: GovAI Chat
• + Added: New commitment (will not): Staff not to use AI to independently provide authoritative advice or make decisions in public facing interactions
• + Added: New commitment (will): Staff to comply with AEC's AI in recruitment guidelines
• ~ Changed: Public-facing AI use: none → unclear
• ~ Changed: Stated last-updated date: 2026-02-09 → 2026-07-31

Adds a "Useful resources" section that links to the DTA's Policy for the responsible use of AI in government, version 2.0. No claims or commitments change.

The ATO adds a policy and legal domain of AI use, now says it complies with version 2.0 of the policy, and says its standalone AI policy has been released.

• Compliance claim moved from version 1.1 to version 2.0 of the AI in government policy
• Standalone staff AI policy now described as released rather than forthcoming
• New policy and legal domain: drafting responses to ministerial correspondence for review and summarising public content for legal and policy research
• Generative AI expansion now covers the analytics for insights usage pattern as well as workplace productivity

• + Added: Domain added: policy and legal
• ~ Changed: Commitment kept (will): Will soon release a standalone AI policy setting staff obligations to support version 2.0

Rebrands as the Tax Ombudsman, adds AI call transcripts and case summaries, and relaxes staff rules so public AI tools may be used in low-risk situations.

• Removed: staff restricted to enterprise AI only; now publicly available AI tools are allowed in low-risk situations without personal, sensitive or protected information
• Staff explicitly permitted to put personal or sensitive information into enterprise AI tools
• New use disclosed: AI generates call transcripts and case summaries in complaint handling
• Now exploring AI for draft content, case studies and summarising text, images, audio and video

• − Removed: Commitment dropped (will not): Staff permitted only enterprise AI; less secure AI tools restricted
• + Added: New commitment (will not): Publicly available AI tools only in low-risk situations without personal, sensitive or protected information
• + Added: New commitment (will): Progressively restrict URLs of AI tools deemed unfit for use

DSS discloses pilots of full Microsoft 365 Copilot and Copilot agents, marks its AI strategy compliant, adds call transcription, and drops its promised KPIs and blocks on public AI tools.

• New pilots disclosed: licensed Microsoft 365 Copilot and Microsoft Copilot Agents/Copilot Studio, including agents DSS builds itself
• New uses: speech-to-text transcription of telephone calls, and redacting information
• Strategic position now compliant: DSS AI Strategy 2026–2028 approved and published to staff
• Chief AI Officer row added, citing the APS AI Plan: CAIO appointed at SES Band 3

• − Removed: Commitment dropped (will not): Technological controls restrict access to publicly available AI tools from the DSS ICT environment
• − Removed: Commitment dropped (will): Every AI use case will have approved KPIs to track outcomes and measure effectiveness
• + Added: Named tool added: Microsoft Copilot Studio
• + Added: New commitment (will not): AI use outside an approved use case is prohibited
• + Added: New commitment (will): Every proposed AI use case must pass risk, legal, privacy and security assessment and approval before use
• + Added: Safeguard added: audit or assurance
• + Added: Strategic position on AI: planned → in place
• ~ Changed: Commitment kept (will): Will document an AI strategic position by 1 June 2026
• ~ Changed: Stated last-updated date: 2026-01-22 → 2026-07-22

Nothing of substance changed: the page title now shows up as a duplicate top-level heading above the existing one, most likely because the page layout or scraper output shifted.

Annual refresh names a Chief AI Officer, moves the chatbot from the Census Test to the 2026 Census with privacy safeguards, and adds staff training requirements and legislative compliance claims.

• Chief AI Officer appointed under the APS AI Plan 2025
• Public chatbot now serves the full August 2026 Census rather than the 2025 Census Test
• Chatbot described as using a fixed knowledge base, not learning from user input, and automatically removing personal details
• Possible future uses flagged, including testing and assurance, limited to no public interaction, personal or market-sensitive information, and subject to staff review

• + Added: Chief AI Officer: not mentioned → in place
• + Added: New commitment (human oversight): Outputs of future AI testing and assurance uses will be subject to review by ABS employees
• + Added: New commitment (will not): Future AI uses for testing and assurance will not involve public interaction, personal information or market sensitive information
• + Added: Safeguard added: acceptable use policy
• + Added: Staff training: available → mandatory
• ~ Changed: Stated first-published date: 2025-07-16 → 2026-07-15

Rewrites its statement to stand on its own rather than lean on PM&C's. It names its CIO as accountable official, discloses its AI uses, adds an annual review, gives its own contact and drops the public-interaction safeguard.

• Removed: commitment not to use AI where the public may directly interact with it or be significantly affected by it without a human intermediary
• Removed: principle that AI must not be the final decision-maker on government advice and services
• Removed: ban on putting classified, sensitive or personal information into AI
• Removed: rule allowing only AI tools authorised on PM&C's ICT systems, and the statement that NZEA runs on PM&C's systems and follows PM&C's AI statement

• − Removed: Commitment dropped (will not): AI input must not include or reveal classified, sensitive or personal information
• − Removed: Commitment dropped (will not): Will not use AI where the public directly interacts or is significantly impacted without a human intermediary
• − Removed: Commitment dropped (will): Only AI tools authorised for access and use within PM&C's ICT systems are deployed
• − Removed: Commitment: no public-facing AI without a human removed
• − Removed: Safeguard dropped: privacy or security controls
• + Added: Accountable official: not mentioned → designated
• + Added: New commitment (will): Centralised oversight and evaluation of AI tools through the Accountable Official and Executive Team
• + Added: New commitment (will): Review statement at least every 12 months and update if AI approach changes
• + Added: Safeguard added: governance body
• + Added: Safeguard added: testing or evaluation
• + Added: Usage pattern added: analytics for insights
• + Added: Usage pattern added: workplace productivity
• ~ Changed: Public-facing AI use: with human review → not addressed
• ~ Changed: Review cadence: not stated → annual and on change

Writes out "Artificial Intelligence (AI)" in full in the opening sentence about the DTA policy's requirements; nothing of substance changed.

Drops a brief introductory note explaining the difference between its strategic position on AI and its AI transparency statement; the substance is unchanged.

No change to the statement itself: the site's Acknowledgement of Country footer and banner image now appear at the end of the captured page.

Changes the statement's most recent update from June to July 2026 and adds a link to its general contact page for enquiries.

Reports that a Chief AI Officer was appointed in June 2026, replacing the earlier promise to appoint one by July, and updates the statement date.

• Chief AI Officer appointed in June 2026, replacing the commitment to appoint one by July 2026
• Chief AI Officer's role described: championing strategic change and driving AI adoption and engagement
• Statement date changed from 7 April 2026 to 3 July 2026

• + Added: Chief AI Officer: planned → in place
• ~ Changed: Commitment kept (will): A Chief AI Officer will be appointed by July 2026, as required by the APS AI Plan
• ~ Changed: Stated last-updated date: 2026-04-07 → 2026-07-03

Healthdirect now names its AI tools. It also discloses two new uses, AI website translation and AI scribes for GPs, and widens its future plans to include commercial AI assistants.

• New use: AI scribes (Heidi or Lyrebird) record GP consultations on 1800MEDICARE (GP)
• New use: Google Translate translates website content, with a disclaimer and quality-assurance safeguards covering consumer feedback, targeted reviews and possibly disabling translations
• Tools named: Infermedica triage, Microsoft Copilot, Claude Enterprise, Adobe Firefly, AWS Kiro
• Infermedica named as the technology partner behind the Symptom Checker

• + Added: Named tool added: Adobe Firefly
• + Added: Named tool added: Claude Enterprise
• + Added: Named tool added: Google Translate
• + Added: Named tool added: Heidi
• + Added: Named tool added: Infermedica
• + Added: Named tool added: Kiro
• + Added: Named tool added: Lyrebird
• + Added: Named tool added: Microsoft Copilot
• + Added: New commitment (will): Address identified AI translation issues and may disable translations under clinical and content governance processes
• ~ Changed: Stated last-updated date: 2025-11 → 2026-06

Confirms a Chief AI Officer is now in place, stops naming the Chief Operating Officer as accountable official, drops its enquiries contact and moves the last-update date to June 2026.

• Chief AI Officer now in place, working with the Accountable Official to drive AI adoption (previously 'will be appointed by July 2026')
• Removed: Chief Operating Officer named as the AI accountable official; the role is now described without naming who holds it
• Removed: enquiries contact via the Chief Operating Officer at services@cgc.gov.au
• Most recent update date changed from March 2026 to June 2026

• − Removed: Public contact removed
• + Added: Chief AI Officer: planned → in place
• + Added: New commitment (will): APSC AI in Government Fundamentals course is mandatory for all staff under the APS AI Plan
• ~ Changed: Commitment kept (will): A Chief AI Officer will be appointed by July 2026 to lead AI transformation
• ~ Changed: Stated last-updated date: 2026-03 → 2026-06

Fixes a malformed enquiry email address, changing communications@www.screenaustralia.gov.au to communications@screenaustralia.gov.au.

Moves the Accountable Officer role from the General Manager to a new information governance officer, names the General Manager Chief AI Officer, and justifies its ban on automated decisions in more detail.

• Accountable Officer changed from the General Manager to the Chief Information Governance and Security Officer
• General Manager named Chief AI Officer under the APS AI Plan
• New Information Governance Committee coordinates management of AI and emerging risks, and reports to the EMT
• Now explicitly rules out decisions under the PSR Scheme being made or influenced by automated systems, citing the Administrative Review Council and the Commonwealth Ombudsman

• + Added: Chief AI Officer: not mentioned → in place
• ~ Changed: Stated first-published date: 2025-02-28 → —
• ~ Changed: Stated last-updated date: — → 2026-02

Names a Chief AI Officer, makes the Chief Operating Officer the sole accountable official, says Copilot is now in use rather than on trial, and adds CTM Scout and AI use classifications.

• Group Executive Director, Systems Assurance and Data Analytics, appointed Chief AI Officer
• Accountable Official narrowed to the Chief Operating Officer; Removed: Chief Digital Officer as an Accountable Official
• Microsoft Copilot moves from trial to adoption across the ANAO and is now rated low risk
• New AI tool disclosed: CTM Scout travel-booking chatbot, exempt from assessment under the ANAO's AI policy

• + Added: Chief AI Officer: not mentioned → in place
• + Added: Named tool added: CTM Scout
• + Added: New commitment (human oversight): AI use in audit activities must be documented in audit planning and subject to testing, validation and review of outputs
• + Added: New commitment (will): Future AI development or customisation subject to additional governance, risk assessment and assurance processes
• + Added: Safeguard added: audit or assurance
• + Added: Safeguard added: human review of outputs
• + Added: Safeguard added: incident or concern reporting
• + Added: Safeguard added: privacy or security controls
• + Added: Safeguard added: testing or evaluation
• + Added: Usage pattern added: analytics for insights
• ~ Changed: Stated last-updated date: 2025-07-21 → 2026-06-26

Updates the linked compliance priorities title from 2025–26 to 2026–27; the AI transparency content is unchanged.

Now says it uses AI in internal service delivery, including case management and handling enquiries, with human review of outputs. It also describes monitoring and drops its promise to develop an AI usage policy.

• AI now used to support internal service delivery, including case management and responding to enquiries from approved information sources
• Removed: statement that WGEA does not use AI in service delivery
• New safeguard: all AI outputs reviewed and approved by staff, advisory only
• New monitoring disclosure: tracks usage levels and outcomes, with reporting for quality assurance

• − Removed: Commitment dropped (will): Will develop a responsible AI usage policy aligned with DTA resources
• + Added: Domain added: service delivery
• + Added: Monitoring and protection measures added
• + Added: New commitment (human oversight): All AI outputs reviewed and approved by WGEA staff before use; AI advisory only
• + Added: New commitment (human oversight): All external communications reviewed and issued by WGEA staff
• + Added: New commitment (will not): AI not used for statutory decision-making, compliance determinations, auditing outcomes or enforcement actions
• + Added: Safeguard added: human review of outputs
• ~ Changed: Public-facing AI use: none → with human review

Fixes a stray reference to the COO, which now reads General Manager, so the accountable officer is the one named as sitting on the executive team; also corrects a typo.

Commits to reviewing and updating its AI transparency statement every year or whenever significant changes occur, filling in a Review and Updates section that had been empty.

• Added: statement to be reviewed and updated annually or when significant changes occur

• + Added: New commitment (will): Review and update the statement annually or when significant changes occur
• ~ Changed: Review cadence: on change → annual and on change

Tidies formatting and capitalisation, moves the DTA classification link into a footnote, and drops the note that the statement was first published in February 2025.

Commits to updating its AI transparency statement as its approach to AI evolves and at least every 12 months.

• Added a review commitment: the statement will be updated as its AI approach evolves and at least every 12 months

• + Added: New commitment (will): Update the statement as the approach to AI evolves, and at least every 12 months
• ~ Changed: Review cadence: not stated → annual and on change

Fills its empty "Review and Updates" section with a promise to review and update the statement every year or whenever significant changes occur.

• Commits to reviewing and updating the statement annually or when significant changes occur

• ~ Changed: Review cadence: annual → annual and on change

Commits to reviewing and updating its AI transparency statement every year, or sooner when significant changes occur.

• Added a review cadence: annual review, or sooner if significant changes occur

• + Added: New commitment (will): Will review and update the statement annually or when significant changes occur
• ~ Changed: Review cadence: not stated → annual and on change

Names a new Chief AI Officer and discloses that it has approved one AI tool, begun limited testing for corporate, policy and legal work, and trained all staff.

• Chief AI Officer changed from Executive Director, Enterprise Management Unit to Assistant Inspector-General, Agency Oversight
• New: one AI tool on the portfolio department's platform approved for use
• New: limited testing and evaluation of AI under way for corporate and enabling, and policy and legal functions
• New: test and evaluation activity logged in centralised documentation

• + Added: Domain added: corporate and enabling
• + Added: Domain added: policy and legal
• + Added: Safeguard added: testing or evaluation
• ~ Changed: Commitment kept (will): Before deployment, all staff who will use AI receive appropriate training

Commits to updating its statement whenever its approach to AI changes, and at least every twelve months.

• New commitment to update the statement at least every twelve months, and whenever its approach to AI changes

• + Added: New commitment (will): Statement will be updated as the approach to AI changes, and at least every twelve months
• ~ Changed: Review cadence: not stated → annual and on change

Adds a commitment to review and update the statement every year, or sooner if significant changes occur, under its previously empty "Review and updates" heading.

• New review cadence: every year, or sooner when significant changes occur

• + Added: New commitment (will): Review and update the statement annually or when significant changes occur
• ~ Changed: Review cadence: annual → annual and on change

Austrade commits to updating its statement at least every twelve months, and whenever its approach to AI changes significantly or new factors affect the statement's accuracy.

• Added: commitment to update the statement at least every twelve months
• Added: commitment to update after significant changes to its approach to AI or when new factors materially affect accuracy

• + Added: New commitment (will): Update statement on significant change to AI approach or accuracy, and at least every twelve months
• ~ Changed: Review cadence: not stated → annual and on change

Adds a commitment to review and update its AI transparency statement annually or when significant changes occur.

• Added review cadence: annual review, or sooner when significant changes occur

• ~ Changed: Review cadence: annual → annual and on change

No change to the statement itself; a "Go to section" navigation heading from the page template dropped out of the capture.

Rewrites its statement for policy v2.0. It now discloses current use of generative AI, including image generation, names a Chief AI Officer appointed in February 2026 and adds a contact email.

• Chief AI Officer appointed in February 2026
• Workplace productivity AI moved from potential benefit to current use of generative AI
• New disclosed uses: image processing and generation, and dataset analysis for patterns and trends
• New commitment that AI will not make final regulatory or enforcement decisions without accountable human judgement

• − Removed: Commitment dropped (will): If plans change on public-facing AI, will update this statement to detail AI use
• − Removed: Commitment dropped (will): Will be transparent in internal use of AI as we explore, evaluate and adopt it
• − Removed: Strategic position on AI: in place → not mentioned
• + Added: Chief AI Officer: not mentioned → in place
• + Added: Legislative compliance statement added
• + Added: New commitment (human oversight): AI supports, never replaces, staff professional judgement; decisions and advice remain with ACMA staff
• + Added: New commitment (human oversight): AI will not make final regulatory or enforcement decisions without accountable human judgment
• + Added: New commitment (will not): Will not use AI inconsistently with legal obligations, public trust responsibilities or Australian Government policy
• + Added: New commitment (will): Any future AI use subject to appropriate governance, risk assessment and human oversight
• + Added: New commitment (will): Will consult staff and representatives on AI uses materially affecting work practices, roles or responsibilities
• + Added: Public contact added
• + Added: Safeguard added: human review of outputs
• + Added: Safeguard added: incident or concern reporting
• + Added: Usage pattern added: decision making and administrative action
• + Added: Usage pattern added: image processing
• ~ Changed: Policy version referenced: v1 → v2
• ~ Changed: Stated last-updated date: 2026-02 → —

The statement is unchanged. The scraper stopped picking up the website's related-page links at the bottom (Learn, Board, APS Census, Careers, Contact us, Volunteer).

Discloses GenAI use for research and analysis support and for drafting publications and briefings, while limiting AI to support functions and making staff responsible for verifying outputs.

• New 'Research and analysis support' use: summarising information, structuring analyses and preparing reports, publications and briefing content
• Workplace productivity use widened from structuring internal reports to structuring and refining publications and briefing materials
• New limit: GenAI confined to support functions and does not replace human authorship, decision-making or accountability
• New safeguard: staff responsible for verifying the accuracy and appropriateness of all AI-assisted outputs

• + Added: New commitment (human oversight): Staff remain responsible for verifying accuracy and appropriateness of all AI-assisted outputs
• ~ Changed: Stated last-updated date: 2026-05-06 → 2026-06-19

Screen Australia's AI enquiries contact email now shows in full where the previous capture had Cloudflare's "[email protected]" placeholder; the statement text itself is unchanged.

Adds a new organisation-wide strategic position on AI adoption ahead of the transparency statement. It also updates the page date and demotes the statement's headings under a new parent section.

• New 'Strategic position on AI adoption' section, framed around quality, confidentiality and accountability in a high-trust parliamentary setting
• Strategic position published as a one-page image, so its detailed content isn't captured as text
• Page date changed from 23 February 2026 to 18 June 2026

• + Added: New commitment (will): COO or delegate informs the Parliamentary AI Working Group and DPS ICT before using new AI
• + Added: Strategic position on AI: not mentioned → in place
• ~ Changed: Stated last-updated date: 2026-02-23 → 2026-06-18

Makes a small wording change to how it describes its workplace productivity use of AI, turning "such as" examples into a flat list. The substance is unchanged.

Discloses that AI was used in evaluating Digital Marketplace Panel 2 applications, gives the AI Branch Manager (replacing the CTO) primary policy responsibility, and logs a 17 June 2026 revision.

• New disclosed use: AI in part of the Digital Marketplace Panel 2 application evaluation, under the AI assurance framework with external probity, privacy and technical advice
• Primary responsibility for cross-government coordination and further policy actions moved from the CTO to the AI Branch Manager
• Change log entry dated 17 June 2026 added (aligned with the transparency statement requirements, accountable official details updated)

• ~ Changed: Stated last-updated date: 2025-12-01 → 2026-06-17

The Tribunal discloses a Microsoft Copilot trial and other internal Microsoft AI pilots, bans public AI, renames its mandatory training and adds a compliance table with December 2026 deadlines.

• New use disclosed: coordinated trial of premium Microsoft Copilot features for staff
• New use disclosed: pilots of other secure Microsoft AI products for back-office work, including software development and cyber security automation
• New safeguard: AI with access to Tribunal data stays in its controlled environment; public AI is prohibited
• Mandatory training renamed to 'AI in Government: Unlocking Innovation with Accountability'

• + Added: AI use-case register: not mentioned → planned
• + Added: Domain added: corporate and enabling
• + Added: Named tool added: Microsoft 365 Copilot
• + Added: New commitment (will not): Access to and use of Public AI is prohibited
• + Added: New commitment (will): Frameworks and processes to operationalise responsible AI due 1 December 2026
• + Added: New commitment (will): Internal AI use case register in development, due 1 December 2026
• + Added: New commitment (will): Process to assess AI use cases with accountable use case owners due 1 December 2026
• + Added: Safeguard added: privacy or security controls
• + Added: Strategic position on AI: not mentioned → in place
• + Added: Usage pattern added: workplace productivity
• ~ Changed: Policy version referenced: unspecified → v2

No substantive change: the page title and site navigation and subscribe links are no longer captured, and bold heading formatting and one space are gone.

Adds new sections on AI risk controls and on monitoring. Describes its AI governance in specific terms, rules out AI in service delivery and corrects "Treasury" to the AOFM.

• New risk controls: AI limited to approved tools in the Treasury ICT environment, and no sensitive, classified or market-sensitive information entered into public AI tools
• New monitoring commitments: initial cost, risk and benefit assessments of AI use cases, oversight through existing forums, and stronger evaluation as AI adoption grows
• Admits its performance measurement frameworks are basic, given limited, low-risk use
• Governance now described as part of existing management, risk and information arrangements, including a specialised AI focus group

• − Removed: Commitment dropped (will): Publish an annual AI Transparency Statement on how AI is used and managed
• + Added: New commitment (will not): No entry of sensitive, classified or market-sensitive information into public AI tools
• + Added: New commitment (will): Continue assessing AI risks (accuracy, bias, data handling) and adapt controls as needed
• + Added: New commitment (will): Restrict AI use to approved tools within the Treasury ICT environment
• + Added: New commitment (will): Strengthen monitoring and evaluation arrangements as AI adoption increases
• + Added: Safeguard added: governance body
• + Added: Safeguard added: risk assessment
• ~ Changed: Stated last-updated date: 2026-02-28 → 2026-06-12

The ALRC now discloses an AI pilot using named legal research tools for inquiry-related work, and drops its description of the staff working group and the staff guidance acknowledgement requirement.

• New pilot uses legal AI tools (LexisNexis Lexis+ AI with Protégé, Westlaw Precision Australia with AI-Assisted Research) for legal research, analysis and synthesis
• AI may now be used for inquiry-adjacent tasks such as analysing submission and consultation material, checking sources and proofreading legal and policy work
• Commits that AI will not be used for legal reasoning, independent analysis, or inquiry findings and recommendations
• States that all AI outputs are reviewed and verified against sources, and staff remain accountable

• − Removed: Commitment dropped (will not): Do not distribute sensitive material to third parties by pasting it into AI software
• − Removed: Commitment dropped (will): Staff must acknowledge the AI use policy before accessing any generative AI tools
• − Removed: Safeguard dropped: privacy or security controls
• + Added: Domain added: policy and legal
• + Added: Named tool added: Lexis+ AI
• + Added: Named tool added: Westlaw Precision Australia
• + Added: New commitment (will not): AI trial not used for legal reasoning, substantive work from scratch, or inquiry findings or recommendations
• + Added: Usage pattern added: analytics for insights

Austrade rewrites its statement for Standard 2.0. It reports Copilot rolled out to all staff, discloses AI social media monitoring, adds a public-interaction commitment and moves AI review to an AI Working Group.

• Copilot moved from a limited trial to all staff
• New disclosure: tools with AI features monitor public digital channels and social media
• New commitment: no AI that the public directly interacts with, or is significantly affected by, without human review
• Mandatory AI Foundations training and an acknowledged AI-use policy now required before using generative AI

• − Removed: Commitment dropped (will): Develop a security risk management plan and ethics assessment for any proposed AI tool
• − Removed: Commitment dropped (will): Update this transparency statement when Austrade approves AI solutions for use
• − Removed: Usage pattern dropped: decision making and administrative action
• + Added: AI use-case register: not mentioned → in place
• + Added: Domain added: service delivery
• + Added: New commitment (human oversight): Staff personally accountable for AI outputs and able to explain and justify decisions based on them
• + Added: New commitment (human oversight): Staff review and verify all AI outputs in full, including for bias and discrimination
• + Added: New commitment (will not): Do not enter sensitive, classified or personally identifiable information into publicly available AI tools
• + Added: New commitment (will): AI use cases assessed for potential negative impacts
• + Added: New commitment (will): Disclose when AI is used in material for clients, partners or the public
• + Added: New commitment (will): Security risk management plan and system authorisation required for each AI tool before deployment
• + Added: New commitment (will): Staff must acknowledge the AI tools policy before accessing generative AI tools online
• + Added: New commitment (will): Staff must complete AI Foundations training before using Copilot
• + Added: Safeguard added: acceptable use policy
• + Added: Safeguard added: staff training
• + Added: Safeguard added: use case register
• + Added: Staff training: not mentioned → mandatory
• ~ Changed: Policy version referenced: unspecified → v2
• ~ Changed: Public-facing AI use: not addressed → none
• ~ Changed: Review cadence: on change → not stated
• ~ Changed: Stated first-published date: — → 2025-02-18
• ~ Changed: Stated last-updated date: — → 2026-06-02

Adds a "(PDF)" label to the link to the DTA's classification system for AI use. Nothing else in the statement changed.

Reports that its Chief AI Officer has now been appointed, stops naming the CIO as accountable official, and moves enquiries to a new AI contact address.

• Chief AI Officer now appointed (was due by July 2026)
• Removed: identification of the Chief Information Officer as the AI accountable official
• Enquiry contact changed from the CIO at ISBCoord@treasury.gov.au to ai@treasury.gov.au
• Most recent update date moved from 28 February 2026 to 3 June 2026

• + Added: Chief AI Officer: planned → in place
• ~ Changed: Commitment kept (will): A Chief AI Officer will be appointed by July 2026 to lead AI transformation
• ~ Changed: Stated last-updated date: 2026-02-28 → 2026-06-03

DFAT names a Chief AI Officer and AI Accountable Officer in place of the CIO, and now requires AI training only of staff who use AI, not before access.

• Chief AI Officer and AI Accountable Officer appointed, in line with whole-of-government requirements
• Removed: the Chief Information Officer as the officer responsible for compliance with whole-of-government AI policy and for departmental data quality and integrity
• Removed: the requirement that staff finish responsible-use training before they are given AI access (now 'all staff using AI must complete training')
• Governance processes now 'support' taking action to reduce harm instead of ensuring it

• + Added: Chief AI Officer: not mentioned → in place
• ~ Changed: Stated last-updated date: 2025-12-03 → 2026-06-03

No change to the statement itself: only the download widget label and file-size text around the data strategy PDF link moved or changed.

No change to MoAD's statement: only the "related pages" tiles in the page footer changed, such as Learn, Board, Careers and Volunteer.

The statement text is unchanged. Only the related-page tiles in the museum website's footer rotated, swapping links such as Board and Careers for Media and Contact us.

No change to the statement itself; only the site's related-links tiles rotated, swapping map, parking and reports links for conditions of entry, volunteering and careers.

Only the "related pages" sidebar tiles below the statement changed on the museum's website. The statement text itself is unchanged.

Nothing in the statement changed. The only differences are in the related-pages tiles below the statement, which were swapped and reordered.

Moves the deadline for all staff to complete the APS Academy's AI in Government Fundamentals course from 31 May 2025 to 30 June 2026.

• Staff AI training deadline moved from 31 May 2025 to 30 June 2026

The statement text is unchanged. Only the "related pages" sidebar tiles below it were reshuffled: venue hire, map and parking tiles now appear, and the careers and collection tiles are gone.

The statement text is unchanged. Only the scraper's rendering of the data strategy PDF link and the "next up" navigation icon differs.

Nothing in the statement changed; the only difference is the related-links tiles at the bottom of MoAD's page, which were reshuffled.

Nothing in MoAD's statement changed; only the page's related-links tiles (volunteer, venue hire, partnerships, media and similar) were reshuffled below the statement.

Tweaks the contact sentence to invite "questions or concerns" about its AI use; the email address and contact point are unchanged.

No change to the statement itself; the museum's related-pages sidebar now shows different link tiles, such as Map and Conditions of entry in place of Board and Partnerships.

Moves its AI transparency statement into a section of its corporate governance page. The move drops the "Why we use AI" section and the last-updated date.

• Statement now sits as a section of the corporate governance page, next to privacy, FOI and PID material
• Removed: 'Why we use AI' section explaining the purpose of AI adoption
• Removed: 'Last updated 5 March 2026' date stamp
• Remaining AI content (Copilot use, human review, no AI decision-making, CIO as accountable official) unchanged

Rewrites its statement: Copilot moves from testing to rollout, with mandatory staff training, a planned AI oversight committee, a policy v2.0 compliance table and new review triggers. The contact email is dropped.

• Microsoft 365 Copilot moves from testing to rollout, and Copilot Chat is added
• Staff must complete generative AI training before using Copilot; an AI use policy that staff must acknowledge is being developed
• An AI oversight committee is being established to govern AI use
• AI tools are limited to internal corporate functions and are not authorised to make decisions or replace human judgement

• − Removed: Public contact removed
• + Added: New commitment (human oversight): Our people will continue to be responsible for oversight and decision making
• + Added: New commitment (will not): AI tools not authorised to independently make decisions, give official determinations or replace human judgement
• + Added: New commitment (will): Develop staff AI-use policy that staff must acknowledge before accessing generative AI tools
• + Added: New commitment (will): Establish an AI oversight committee responsible for governance, risk, human oversight and monitoring
• + Added: New commitment (will): Staff must complete internal generative AI training before using Copilot
• + Added: Safeguard added: acceptable use policy
• + Added: Safeguard added: audit or assurance
• + Added: Safeguard added: governance body
• + Added: Safeguard added: human review of outputs
• + Added: Safeguard added: privacy or security controls
• + Added: Safeguard added: staff training
• + Added: Safeguard added: testing or evaluation
• + Added: Staff training: not mentioned → mandatory
• ~ Changed: Policy version referenced: v1 → v2
• ~ Changed: Stated first-published date: 2025-01 → —
• ~ Changed: Stated last-updated date: — → 2026-05

Removes the "28 February 2026" date stamp from the top of the statement. The statement text itself is unchanged.

No change to the statement itself; only the related-page tiles at the foot of the page were swapped.

Only the page's related-links tiles changed. Careers, contact and similar links were swapped for reports, venue hire, media and map links, and the statement text is untouched.

Only the museum's "related pages" link tiles below the statement changed. The statement text itself is untouched.

No change to the statement itself; only the related-page navigation tiles at the foot of the page were shuffled and swapped.

The statement text is unchanged. Only the "related pages" tiles in the sidebar rotated, swapping venue hire, media and similar links for map, volunteer and contact links.

Nothing in the statement changed: only the related-links tiles under the page were reshuffled, with Venue hire, Conditions of entry, the APS Census and About Old Parliament House swapped in.

Geoscience Australia rewrote its statement for policy v2.0. It cuts its accountable officials from two to one (the CIO), commits to appointing a Chief AI Officer by July 2026, and adds mandatory training.

• Accountable officials cut from two (CIO and Chief Scientist) to the CIO alone
• Commits to appointing a Chief AI Officer by July 2026 under the APS AI Plan
• New requirement for staff to complete mandatory training before using enterprise AI tools
• Names Microsoft 365 Copilot as its endorsed enterprise AI tool

• + Added: Chief AI Officer: not mentioned → planned
• + Added: New commitment (will): A Chief AI Officer will be appointed by July 2026 to lead AI transformation
• + Added: New commitment (will): Staff must complete mandatory training before using Enterprise AI tools
• + Added: Safeguard added: risk assessment
• + Added: Staff training: available → mandatory
• ~ Changed: Policy version referenced: unspecified → v2
• ~ Changed: Stated first-published date: 2025-02-20 → —
• ~ Changed: Stated last-updated date: — → 2026-05-15

No change to the statement itself; only the rotating "related pages" tiles at the foot of MoAD's page were reshuffled (Partnerships, Careers, Volunteer, Learn, Media and others).

No change to the statement itself; only the site's related-pages tiles below it (media, venue hire, careers, partnerships and similar) were swapped.

Adds an AI classification under the DTA Standard: workplace productivity use in corporate and enabling functions. States the public does not interact with its AI, and no AI outputs affect the public without human review.

• New AI classification section citing the DTA transparency statement Standard
• Usage pattern disclosed: workplace productivity, e.g. document summarisation and content creation
• Domain disclosed: corporate and enabling (HR, finance, communication, IT)
• New statement that the public does not directly interact with its AI systems, and no AI outputs significantly affect the public without human review

• + Added: Domain added: corporate and enabling
• + Added: Intentions behind AI use added
• + Added: Usage pattern added: workplace productivity
• ~ Changed: Public-facing AI use: not addressed → none

No change to MoAD's statement. The related-pages tiles at the foot of the page rotated: Volunteer, Board and APS Census were swapped for Media, Venue hire and Contact us.

Rewrites its statement: it now reports internal AI use for productivity, search, coding and research, lists what AI is excluded from, and adds training, safeguards and a contact address.

• Now discloses internal AI use for productivity, information discovery, data analytics and coding, and training and research
• Removed: statement that the ATSB does not use AI to deliver any services
• Removed: mention of taking part in the Microsoft 365 Copilot trial
• New exclusions: no AI in investigative findings, HR decisions or public-facing service delivery

• − Removed: Commitment dropped (will): Will update this statement if it begins using AI in public-facing services
• − Removed: Policy compliance statement removed
• − Removed: Safeguard dropped: governance body
• − Removed: Strategic position on AI: planned → not mentioned
• + Added: Domain added: corporate and enabling
• + Added: Domain added: policy and legal
• + Added: Legislative compliance statement added
• + Added: New commitment (human oversight): AI enhances, not replaces, human decision-making; critical decisions reviewed and approved by qualified officials
• + Added: New commitment (human oversight): All AI outputs verified by a staff member before official use; work owned by a human
• + Added: New commitment (will not): AI not used for HR recruitment, hiring, performance management or areas with direct human impact
• + Added: New commitment (will not): Does not use AI for direct service delivery or as primary contact for external stakeholders
• + Added: New commitment (will not): Does not use AI to make decisions or findings in transport safety investigations
• + Added: New commitment (will): Will regularly update this statement on AI initiatives and their outcomes
• + Added: Public contact added
• + Added: Safeguard added: human review of outputs
• + Added: Safeguard added: privacy or security controls
• + Added: Safeguard added: risk assessment
• + Added: Safeguard added: staff training
• + Added: Staff training: not mentioned → mandatory
• + Added: Usage pattern added: analytics for insights
• + Added: Usage pattern added: workplace productivity
• ~ Changed: Commitment kept (will): Will establish an AI Steering Committee to assess AI opportunities and risks
• ~ Changed: Policy version referenced: v1 → not referenced
• ~ Changed: Stated first-published date: 2025-03 → —

No change to the statement: only the rotating "related pages" navigation tiles at the foot of the page changed.

Removes the "Published 27 March 2025" date line from the top of the statement and leaves the rest of the text unchanged.

Nothing in the statement changed. The only difference is in the website's related-links tiles below it, which were swapped and reordered.

Only the museum's rotating "related pages" tiles changed below the statement; the AI transparency text itself is untouched.

No change to the statement itself; the only difference is the rotating "related pages" tiles (careers, contact, volunteering, board) in the site's footer.

Names Brendan Moon as Chief AI Officer and sets out his duties, drops Matthew Johnston as a second AI Accountable Officer, and leaves Kerryn Kovacevic as the sole AI Accountable Officer.

• Chief AI Officer appointed: Brendan Moon, First Assistant Secretary, Child Care Integrity Division
• Chief AI Officer's duties set out: leading AI adoption, raising AI literacy, sponsoring high-value use cases, reporting to the Executive Board within agreed risk limits
• Removed: Matthew Johnston (First Assistant Secretary, Strategy, Data and Measurement) as an AI Accountable Officer
• Removed: commitment that a Chief AI Officer would be in place by June 2026 (now filled)

• + Added: Chief AI Officer: planned → in place
• ~ Changed: Commitment kept (will): Chief AI Officer will be in place by June 2026

Only the page's rotating "related pages" tiles changed. The museum's AI transparency statement text is unchanged.

A "Quick Links" sidebar linking to version 1.1 of the DTA AI policy no longer appears after the statement; the statement text itself is unchanged.

Rewrites its statement for policy v2.0. It discloses new Copilot uses in coding, HR and IT agents and research-topic analytics, allows public GenAI tools, and requires AI training for all staff.

• New use: generating and debugging code for systems development and maintenance
• New use: SharePoint agents supporting internal HR and IT advice
• New use: analytics, including classifying research topics internally
• Staff may now use certain public GenAI tools under DTA guidance

• + Added: Usage pattern added: analytics for insights
• ~ Changed: Policy version referenced: unspecified → v2
• ~ Changed: Stated last-updated date: 2025-02-27 → 2026-05-06

Discloses current use of Claude and Microsoft Copilot for internal admin, sets an annual review by the AI Accountable Official, and replaces its long list of planned governance policies with specific scope limits.

• Discloses current use of authorised AI tools, naming Anthropic's Claude and Microsoft Copilot, for summarising and taking minutes
• Adds example uses: summarising reports and feedback, drafting documents, background research
• Adds out-of-scope exclusion: sensitive data on duplicate payments is not to be used with AI until tools are approved for it
• AI Accountable Official to review the statement at least once a year and to oversee use cases and compliance

• − Removed: Commitment dropped (will not): AI-assisted outputs will not be used for decision making purposes
• − Removed: Commitment dropped (will): Develop internal policies on AI governance, acceptable use, privacy, security, risk and required training
• − Removed: Commitment dropped (will): Ensure conformance with proposed mandatory guardrails for AI in high-risk settings
• − Removed: Staff training: mandatory → not mentioned
• + Added: AI use-case register: not mentioned → planned
• + Added: Domain added: corporate and enabling
• + Added: Named tool added: Claude
• + Added: Named tool added: Microsoft Copilot
• + Added: New commitment (will not): Duplicate-payment integrity data excluded from any AI use until tools are approved for such sensitive information
• + Added: New commitment (will): AI Accountable Official will review the transparency statement at least once a year
• + Added: New commitment (will): Continue to evaluate and update data governance requirements against the application of AI
• + Added: New commitment (will): Develop a strategic position on NHFB's AI adoption
• + Added: New commitment (will): Meet DTA policy requirements, including an AI use case register, AI impact assessments and staff guidance
• + Added: Safeguard added: acceptable use policy
• + Added: Safeguard added: privacy or security controls
• + Added: Strategic position on AI: not mentioned → planned
• + Added: Usage pattern added: workplace productivity
• ~ Changed: Commitment kept (will): Adopt Shared Services provider AI policy and guidance, aligned with the DTA's Policy
• ~ Changed: Review cadence: on change → annual and on change

Nothing in the statement changed; only the website's "related pages" tiles below the statement were reshuffled, with Careers, Contact and About tiles swapped for Volunteer, Collection and Reports.

Nothing in the statement itself changed. The only difference is the rotating "related pages" tiles below it, which now show Careers, Partnerships and Media links instead of others.

Rewrites its statement: it now discloses AI summaries of complaints for allocation and triage, and AI form-filling. It also adds prohibited uses and risk-based approval, and drops the annual review commitment.

• New use disclosed: AI summaries of complaints to support allocation and triage, never used to decide on action
• New use disclosed: AI pre-fills complaint forms from letters and emails, confirmed by an officer
• Prohibits AI for statutory decisions and for writing external communications to complainants
• Classifies use under the DTA's workplace productivity pattern and the service delivery and corporate domains

• − Removed: Commitment dropped (human oversight): Each AI use case and technology requires Ombudsman approval after ITGC endorsement
• − Removed: Commitment dropped (will): AI coding assistants used only for approved use cases; all AI-assisted code tested for accuracy before implementation
• − Removed: Commitment dropped (will): Statement reviewed annually or when AI use cases and usage change
• − Removed: Commitment dropped (will): Will be transparent in assessment, adoption and monitoring of AI technology
• − Removed: Named tool dropped: Azure AI Text to Speech
• − Removed: Safeguard dropped: privacy or security controls
• − Removed: Safeguard dropped: testing or evaluation
• − Removed: Staff training: mandatory → available
• + Added: Chief AI Officer: not mentioned → in place
• + Added: Domain added: service delivery
• + Added: New commitment (human oversight): AI-populated complaint form suggestions must be confirmed by an Officer before recording
• + Added: New commitment (human oversight): Each use case needs approval after impact assessment; Chief AI Officer for low/medium risk, Executive Committee for high risk
• + Added: New commitment (human oversight): Statutory decisions always made by an Officer with the appropriate delegation
• + Added: New commitment (will not): Complaint summaries never used to decide what action to take on a complaint
• + Added: New commitment (will not): Staff not allowed to use AI to generate text for communications outside the Office
• + Added: Safeguard added: human review of outputs
• + Added: Safeguard added: risk assessment
• ~ Changed: Policy version referenced: v1 → v2
• ~ Changed: Public-facing AI use: not addressed → with human review
• ~ Changed: Review cadence: annual and on change → on change
• ~ Changed: Stated first-published date: 2025-02-27 → 2025-02
• ~ Changed: Stated last-updated date: 2025-12-20 → 2026-04

Only the museum's "related pages" sidebar tiles below the statement changed. Careers, Learn, Volunteer and APS Census were swapped for Reports, Board, Conditions of entry and About links.

Only the page's "related pages" link tiles below the statement changed. The statement text itself is untouched.

Nothing in the statement changed; only the rotating "related pages" tiles at the foot of the page swapped links such as Volunteer and Learn for Venue hire and Parking.

Rewrites its statement: names its Chief Information Officer as accountable official plus a Chief AI Officer, discloses cyber-security and research uses, and commits to human oversight without public interaction.

• Names the Chief Information Officer as accountable official and appoints a Chief AI Officer
• New disclosed use: anomaly detection in system activity for cyber security, with alerts for human review
• New uses: summarising public reports, research support, simplifying content, analytical support on options and risks
• New commitment: AI is limited to non-decision-making work, does not directly interact with the public, and outputs get mandatory human review

• + Added: Chief AI Officer: not mentioned → in place
• + Added: Commitment: no public-facing AI without a human added
• + Added: New commitment (human oversight): AI outputs potentially impacting the public subject to mandatory human oversight and approval before action
• + Added: New commitment (human oversight): Staff review and verify AI outputs before use; AI does not replace human judgement
• + Added: New commitment (will not): AI use confined to non-decision-making activities and does not directly interact with the public
• + Added: New commitment (will not): Restrict AI use to approved tools; public AI tools only with information already public
• + Added: New commitment (will): Any expansion of AI use subject to risk assessment, governance approval and policy compliance
• + Added: Safeguard added: human review of outputs
• + Added: Safeguard added: incident or concern reporting
• + Added: Staff training: available → mandatory
• ~ Changed: Public-facing AI use: not addressed → with human review

The Mint's statement text hasn't changed. This capture appears to render it as a web page rather than a PDF, so page numbers and running headers drop out and headings and bullets are reformatted.

The same statement text, now taken from a web page rather than a PDF, which drops the page footers and adds Markdown headings and bullets; the content is unchanged.

Republishes the unchanged December 2025 statement with proper headings and bullet formatting, apparently moving it from a PDF-style layout to a web page; the wording is identical.

The Fair Work Ombudsman's statement text is unchanged. The capture has moved from a PDF-style layout to clean Markdown headings and bullets, which drops page footers and PDF hyphenation artefacts.

The Bureau of Meteorology's statement is unchanged. The text now comes through with proper headings and bullets instead of raw PDF-style line breaks.

The Museum's April 2025 statement is unchanged; this capture only converts the same text differently, with headings, rejoined lines, bullet markers and no page numbers.

Same statement text captured in a different layout, with PDF page headers dropped and Markdown headings and bullets added. The wording and commitments are unchanged.

Reformats headings, italics and list spacing, updates link targets for the DTA policy and Terms of Trade, and adds links to its AI Guiding Principles and the DTA classification system.

Only the related-links tiles at the foot of the page changed. The statement text is untouched.

Austrade's statement text is unchanged. The only addition is a site-wide banner pointing businesses to the Go Global Toolkit for supply chain disruption updates.

Nothing of substance changed: Cloudflare's email protection hid the ALRC's contact address (info@alrc.gov.au), so the capture shows "[email protected]" in its place.

Rewrites its statement, naming new AI officials and adding AI use for corporate functions. It drops its pledge to keep AI away from direct public interaction and its DTA compliance claim.

• Chief Information Officer appointed AI Accountable Official, replacing the Group Manager, Corporate & Engagement
• Chief Operating Officer appointed Chief AI Officer
• New AI use disclosed: support for corporate and enabling functions
• Staff must now complete mandatory training before getting AI access (previously training was only available)

• − Removed: Commitment dropped (will not): Do not propose to use AI where the public may directly interact with or be significantly impacted by it
• − Removed: Commitment dropped (will): Adapt AI systems and advice to align with evolving legislation, ethical standards and public expectations
• − Removed: Commitment dropped (will): Any changes to AI use will result in an update to the transparency statement
• − Removed: Commitment dropped (will): Release an annual AI transparency statement on how AI is used and managed
• − Removed: Commitment dropped (will): Will only use AI services in accordance with applicable legislation, regulations, frameworks and policies
• − Removed: Commitment: no public-facing AI without a human removed
• − Removed: Legislative compliance statement removed
• − Removed: Policy compliance statement removed
• − Removed: Safeguard dropped: incident or concern reporting
• + Added: Chief AI Officer: not mentioned → in place
• + Added: Domain added: corporate and enabling
• + Added: New commitment (will): Staff must complete mandatory training before gaining access to AI
• + Added: Staff training: available → mandatory
• ~ Changed: Policy version referenced: unspecified → not referenced
• ~ Changed: Public-facing AI use: none → not addressed
• ~ Changed: Stated first-published date: 2025-02-28 → —

Comcare now says it is exploring limited internal AI use and running a governance program aligned to DTA guidance, and records an April 2026 review.

• New disclosure: exploring AI in a limited, controlled way for internal analysis, information management and workplace productivity
• New commitment: targeted program aligned to DTA guidance to build a risk-based framework for identifying, governing and overseeing AI use
• Focus stated as governance, assurance and visibility of AI activity rather than delivering individual AI solutions
• Statement now records a review in April 2026

• + Added: Domain added: corporate and enabling
• + Added: New commitment (will): Establish a practical, risk-based framework for identifying, governing and overseeing AI use across the agency
• + Added: Strategic position on AI: not mentioned → planned
• + Added: Usage pattern added: analytics for insights
• + Added: Usage pattern added: workplace productivity
• ~ Changed: Stated last-updated date: — → 2026-04

Tidies punctuation and italics in the AI statement without changing substance; the surrounding website privacy notice now also covers ausalert.gov.au.

Points to a new 2026 policy on generative AI in ARC grant programs, which covers all programs and applicants, administering organisations and assessors, and takes effect 28 April 2026.

• Links a new 2026 version of the ARC's policy on generative AI in grant programs, replacing the link to the earlier policy
• New policy applies to all ARC grant programs and guides applicants, administering organisations and assessors
• States the policy takes effect from 28 April 2026

Says the statement was last reviewed in February 2026 and widens its review triggers to include changes in the Government's approach to AI, with review at least annually.

• Discloses a review of the statement in February 2026 (previously only 'first published February 2025')
• Review trigger extended to significant changes in the Government's approach to AI, not just the ACMA's
• Review cadence now 'at least annually' rather than 'annually'

• ~ Changed: Stated first-published date: 2025-02 → —
• ~ Changed: Stated last-updated date: — → 2026-02

Rewrites its statement. It now names the Chief Information Officer as Accountable Official and Microsoft Copilot as a tool in use, drops the annual review commitment, and softens the staff training pledge.

• Chief Information Officer named as the Accountable Official for AI oversight and risk management
• Microsoft Copilot named as a current AI tool, managed through existing ICT, security and risk frameworks
• Removed: commitment to review the statement annually or when significant changes occur
• Training commitment weakened from all staff 'will receive' training to all staff being given 'access to' training

• − Removed: Commitment dropped (will): All staff will receive training on the responsible use of AI
• − Removed: Commitment dropped (will): Committed to keeping stakeholders informed about AI initiatives
• − Removed: Commitment dropped (will): Will review this statement annually or whenever significant changes occur in approach to AI
• − Removed: Staff training: mandatory → available
• + Added: Accountable official: not mentioned → designated
• + Added: Named tool added: Microsoft Copilot
• + Added: New commitment (will): Higher-risk AI uses subject to stronger controls and oversight under a risk-based approach
• + Added: Safeguard added: privacy or security controls
• + Added: Safeguard added: risk assessment
• ~ Changed: Public-facing AI use: with human review → none
• ~ Changed: Review cadence: annual and on change → on change

Adds a commitment that no public-facing AI runs without human oversight, classifies its AI use under the DTA scheme, and rewrites its data-safety controls.

• New statement that MoAD does not use AI in services where the public interacts directly with AI, or is significantly affected by it, without human oversight
• New 'Classification of AI use' section: usage patterns are workplace productivity and analytics/reporting; domain is corporate and enabling functions
• Data-safety controls now cite periodic reviews, audits and performance evaluation of AI use cases, and escalation thresholds with human review
• Removed: statement that all AI use is subject to strict change and risk management processes to monitor the impact of changes and new features

• + Added: Safeguard added: testing or evaluation
• ~ Changed: Public-facing AI use: not addressed → none

Reissues its statement for 2026, commits to naming a Chief AI Officer by July 2026, cites the APS AI Plan, and reports AI coding, design and testing tools now in use.

• AI-based coding, design and testing automation tools now in use in the IT section (previously a testing tool was only under consideration)
• Commits to appointing a Chief AI Officer by July 2026, as the APS AI Plan requires
• Now cites the AI Plan for the Australian Public Service alongside the Policy
• Statement date and preparation date updated from 1 April 2025 to 7 April 2026

• + Added: Chief AI Officer: not mentioned → planned
• + Added: New commitment (will): A Chief AI Officer will be appointed by July 2026, as required by the APS AI Plan
• ~ Changed: Stated last-updated date: 2025-04-01 → 2026-04-07

Rewrites the statement. It commits to appointing a Chief AI Officer by July 2026, makes AI training a prerequisite for staff, and narrows its wording on public-facing AI.

• Chief AI Officer to be appointed by July 2026 to lead AI transformation under the APS AI Plan
• Staff must complete internal AI training before using AI tools, and the APSC AI in Government Fundamentals course is mandatory
• Public-interaction wording changes from not being 'significantly impacted by' AI without human intervention to 'does not currently use' AI making decisions affecting individuals without human oversight
• New commitment that AI use will be environmentally considerate and in the public interest

• − Removed: Commitment dropped (will): Impacted staff will have access to training aligned with responsible and acceptable use of AI policies
• + Added: Chief AI Officer: not mentioned → planned
• + Added: New commitment (will): A Chief AI Officer will be appointed by July 2026 to lead AI transformation
• + Added: New commitment (will): Staff must complete internal AI training as a prerequisite to using AI tools
• + Added: Staff training: available → mandatory
• ~ Changed: Stated last-updated date: — → 2026-03

ACIAR rewrites its statement for policy v2.0. It adds governance frameworks, privacy and security, human review, a contact point and an annual review cadence, and drops its detailed staff-use rules.

• Aligns with policy v2.0, the transparency statement Standard and staff guidance on public generative AI tools
• New commitment: all AI outputs reviewed by staff before use; staff responsible for final decisions
• Commitment changed from 'AI not used for decision-making' to 'no automated decisions directly affecting the public, funding recipients or program beneficiaries'
• Adds governance references: AI Impact Assessment Tool, technical standard, procurement guidance, APS AI Plan 2025

• − Removed: Commitment dropped (will not): AI tools must not access restricted information, perform illegal activities, or assess staff performance
• − Removed: Commitment dropped (will not): Email accounts not used with public AI services without approval
• − Removed: Commitment dropped (will not): Official information must not be entered into public AI services unless already public
• − Removed: Commitment dropped (will): AI technologies adopted only after thorough risk assessment and monitored for compliance
• − Removed: Commitment dropped (will): Administrative decisions aided by AI will be clearly indicated
• − Removed: Commitment dropped (will): Automated decision-making tools need significant risk assessment and approval
• − Removed: Commitment dropped (will): Commercial AI tools require executive approval
• − Removed: Commitment dropped (will): Continuous monitoring so AI policies, tools and usage remain current, overseen by internal governance mechanisms
• − Removed: Commitment dropped (will): Misuse or identified risks must be reported immediately
• − Removed: Safeguard dropped: acceptable use policy
• − Removed: Safeguard dropped: incident or concern reporting
• + Added: New commitment (human oversight): AI tools support staff productivity and do not replace human judgement; staff remain responsible for final decisions
• + Added: New commitment (will not): AI systems do not make automated decisions directly affecting the public, funding recipients or beneficiaries
• + Added: New commitment (will): Statement reviewed annually or when ACIAR's use of AI changes significantly
• + Added: Public contact added
• + Added: Safeguard added: staff training
• + Added: Staff training: not mentioned → available
• ~ Changed: Policy version referenced: unspecified → v2
• ~ Changed: Public-facing AI use: not addressed → none
• ~ Changed: Review cadence: not stated → annual and on change
• ~ Changed: Stated first-published date: — → 2025-03
• ~ Changed: Stated last-updated date: — → 2026-03

Discloses a limited, opt-in trial of AI in recruitment in which humans make every selection decision, and reports that its AI Risk Appetite Guidelines are now finalised.

• New use disclosed: limited opt-in trial of AI in recruitment, used to help compare and evaluate recruitment processes alongside human assessment
• Commits that AI makes no selection decisions: a human panel does all shortlisting, assessment and selection, in line with APS merit principles
• AI Risk Appetite Guidelines changed from 'will finalise' to 'have finalised'
• Last-updated date moved from February 2026 to April 2026

• + Added: New commitment (human oversight): All shortlisting, assessment and selection decisions made by a human selection panel
• + Added: New commitment (will not): Do not use AI to make selection decisions about candidates in recruitment
• ~ Changed: Commitment kept (will): Finalise AI Risk Appetite Guidelines to support AI risk mitigation
• ~ Changed: Stated last-updated date: 2026-02 → 2026-04

ASQA's statement text is unchanged; the capture lost the breadcrumb trail and title, repaired a broken policy link, and swapped a share widget for other page chrome.

Names its Chief Data and AI Officer as the Tribunal's Chief AI Officer and makes that officer jointly responsible for AI approvals with the CIO.

• Chief AI Officer role added: the Chief Data and AI Officer (CDAIO)
• The CDAIO now works with the CIO through the governance framework that approves AI use

• + Added: Chief AI Officer: not mentioned → in place

Fixes a missing comma and "and" in the sentence describing staff use of AI. The substance of the statement is unchanged.

Rewrites its statement: swaps one AI Accountable Officer, promises a Chief AI Officer by June 2026, rules out automated decision-making, but drops its no-public-chatbot commitment and use-case examples.

• Removed: commitment that the public will not interact with a chatbot or similar AI agent
• Removed: ban on fully automated AI outputs and on combining AI with existing ADM programs, replaced by a flat statement that AI is not used for automated decision-making and all decisions are made by staff
• AI Accountable Officer Hamish McDonald replaced by Matthew Johnston (FAS Strategy, Data and Measurement)
• Chief AI Officer to be in place by June 2026

• − Removed: Commitment dropped (will not): Does not allow AI for fully automated output processes
• − Removed: Commitment dropped (will not): Public will not interact with a chatbot or similar AI agent when contacting the department
• − Removed: Commitment dropped (will): Establishing review structures for new uses of AI
• − Removed: Commitment: no public-facing AI without a human removed
• + Added: AI use-case register: planned → in place
• + Added: Chief AI Officer: not mentioned → planned
• + Added: New commitment (human oversight): All decisions are made by staff, ensuring human oversight and accountability
• + Added: New commitment (will): All staff required to complete AI in Government Fundamentals training every 12 months
• + Added: New commitment (will): Chief AI Officer will be in place by June 2026
• + Added: New commitment (will): Developing an AI Strategy and Roadmap for the coming 12–18 months
• + Added: Safeguard added: acceptable use policy
• + Added: Safeguard added: risk assessment
• + Added: Safeguard added: staff training
• + Added: Staff training: not mentioned → mandatory
• + Added: Strategic position on AI: not mentioned → planned
• ~ Changed: Commitment kept (will): Developing an AI use case register as recommended by the Policy
• ~ Changed: Public-facing AI use: none → with human review
• ~ Changed: Review cadence: annual → annual and on change

Changes the statement's stated publication date from 27 February 2025 to 10 April 2026 and leaves the rest of the text unchanged.

Replaces its claim to adhere to the Voluntary AI Safety Standard's 10 guardrails with alignment to the newer Guidance for AI Adoption's 6 essential practices.

• Now says its AI use aligns to the Guidance for AI Adoption (6 essential practices)
• Removed: claim that its AI adoption adheres to the Voluntary AI Safety Standard's 10 guardrails
• Wording softened from 'adheres to' to 'aligns to'
• Date updated changed from 26 February 2026 to 7 April 2026

• − Removed: Commitment dropped (will): Own AI adoption adheres to the Voluntary AI Safety Standard's 10 guardrails
• + Added: New commitment (will): Own AI adoption aligns to the Guidance for AI Adoption's 6 essential practices
• ~ Changed: Stated last-updated date: 2026-02-26 → 2026-04-07

Spells out "Artificial Intelligence (AI)" in the statement's heading and leaves the rest of the statement unchanged.

No change to the Fair Work Commission's statement: the scraper no longer picks up the page's feedback form and star-rating widget.

No change to the statement itself; a contact icon image and the site's page-feedback form dropped out of the capture.

Nothing Comcare wrote has changed: the page's navigation sidebar and share widget are gone, and curly apostrophes have become straight ones.

The only change is a stray "Search search" line from the site's search widget, which has now dropped out of the capture. Austrade's statement text is unchanged.

The ABS statement didn't change; the site's "Search ABS" header text just dropped out of the capture.

Updates the statement for the revised DTA policy, disclosing licensed ChatGPT use, and adds a training prerequisite for access and a requirement to declare AI use when submitting work.

• New AI use disclosed: licensed ChatGPT for some staff, mainly in research areas
• Completing AI training is now required before staff can access AI tools
• Staff must describe their AI use and confirm outputs were reviewed when submitting work for review or approval
• Microsoft 365 Copilot is now available to all staff

• + Added: Named tool added: ChatGPT
• + Added: New commitment (will): Completion of appropriate-use AI training is a prerequisite for access to AI tools
• + Added: Safeguard added: staff training
• + Added: Staff training: not mentioned → mandatory
• ~ Changed: Policy version referenced: unspecified → v2
• ~ Changed: Stated last-updated date: — → 2026-04-02

States it has now appointed a Chief AI Officer, where it previously said the designation was still being formalised, and updates the last-updated date.

• Chief AI Officer now reported as appointed (previously 'being formalised')

• + Added: Chief AI Officer: planned → in place
• ~ Changed: Commitment kept (will): Will formalise designation of a Chief AI Officer
• ~ Changed: Stated last-updated date: 2026-04-01 → 2026-04-02

Rewrites its statement against the DTA classification system, names new governance bodies and an annual review, and swaps its AI email contact for a web form.

• Now classifies its AI use under DTA usage patterns, including decision-support/administrative processes and analytics for insights
• New governance disclosed: AI Governance Working Group, communities of practice, internal AI specialists and CIO/CISO oversight
• Commits to reviewing the statement at least annually; states it was last updated March 2026
• Names 'AI Fundamentals' as the mandatory training for staff using AI tools

• − Removed: Commitment dropped (will): Be clear if AI becomes part of a decision-making process
• − Removed: Commitment dropped (will): Maintain an internal responsible AI usage policy aligned with DTA and APSC guidance
• − Removed: Commitment dropped (will): Provide internal support and means for staff to report AI concerns or queries
• − Removed: Domain dropped: service delivery
• − Removed: Safeguard dropped: acceptable use policy
• − Removed: Safeguard dropped: audit or assurance
• − Removed: Safeguard dropped: incident or concern reporting
• + Added: New commitment (will): All staff accessing AI tools must complete AI Fundamentals mandatory training
• + Added: New commitment (will): Assess training needs as ARPANSA's use of AI evolves
• + Added: New commitment (will): Review statement at least annually and update if approach to or use of AI changes
• + Added: Safeguard added: governance body
• + Added: Usage pattern added: decision making and administrative action
• ~ Changed: Review cadence: not stated → annual and on change
• ~ Changed: Stated first-published date: 2025-02-14 → —
• ~ Changed: Stated last-updated date: — → 2026-03

Names a Chief AI Officer, reports Copilot moved from trial to enterprise use, flags possible AI in grant decision-making, and adds new governance commitments and contacts.

• Chief AI Officer named: Anthony Murfett, Deputy CEO
• Microsoft Copilot 365 Enterprise now in use for meeting transcription and summaries, where it was previously a trial
• Decision-making: moved from 'not currently applied' to possibly using AI to support administrative decisions in grants administration, with human oversight
• Considering AI for grant administration and for coding internal systems

• + Added: Chief AI Officer: not mentioned → in place
• + Added: New commitment (human oversight): Any AI implementation will have risk management practices and ensure human oversight
• + Added: New commitment (will): Introduce an internal prioritisation process to consider new AI projects
• + Added: New commitment (will): Release a strategic position on AI adoption to guide ARC personnel
• + Added: Strategic position on AI: not mentioned → planned
• ~ Changed: Stated last-updated date: — → 2026-03

DVA reports that its consent-based Claims Document Reader AI trial has started rather than being in development, and clarifies that its Semantic Text Analysis is advanced keyword matching.

• Claims Document Reader trial moved from 'developing a trial' to 'has commenced a trial'
• Semantic Text Analysis described as 'an advanced form of key word matching'
• Last updated date changed from 9 February to 1 April 2026

• ~ Changed: Stated last-updated date: 2026-02-09 → 2026-04-01

Adds three AI use domains beyond corporate work, says AI is used to research and draft internal documents, and reissues the statement with a March 2026 date of effect.

• New AI use domains disclosed: Service Delivery, Compliance and fraud detection, and Policy and legal
• New disclosure that the Mint uses AI to develop internal documents, including research and drafting
• Date of effect changed from 6 March 2025 to 25 March 2026
• Approver Emily Martin now listed as CEO, no longer acting CEO

• + Added: Domain added: compliance and fraud detection
• + Added: Domain added: policy and legal
• + Added: Domain added: service delivery
• ~ Changed: Stated last-updated date: — → 2026-03-25

AUSTRAC made minor page edits: it dropped the title heading and page ID, and its generic legal disclaimer now says "certain Australian legislation" instead of "the Act".

ASQA narrows its no-AI claim from all service delivery to public-facing services only, promises privacy, ethics and security assessments for future public-facing AI, and drops its annual review commitment.

• Claim of no AI narrowed from 'service delivery' to 'public-facing services', leaving internal service-delivery use open
• New commitment: privacy, ethics and security assessments as public-facing AI is deployed
• Signals plans to deploy public-facing AI capability
• Removed: commitment to review and update the statement annually or when significant changes occur

• − Removed: Commitment dropped (will): Statement reviewed and updated annually or when significant changes occur
• + Added: Legislative compliance statement added
• + Added: New commitment (will): When deploying public-facing AI, include privacy, ethics and security assessments to protect the public
• ~ Changed: Review cadence: annual and on change → annual

Adds a comma to the Accountable Official sentence. A "Did you find what you were looking for?" feedback widget no longer appears on the page.

Replaces its 2025 statement with a 2026 version that names the CIO as accountable official, rules out AI decision-making, and adds impact-assessment, register and contact commitments.

• Chief Information Officer named as accountable official
• New commitment: NIAA does not use AI for decision-making; all AI outputs reviewed by authorised staff
• Tool now described as Microsoft 365 Copilot Chat for drafting correspondence and summarising meetings
• Will set up an AI impact assessment approach aligned to the DTA tool, covering privacy, security and cultural safety risks

• − Removed: Commitment dropped (will): Abide by PM&C ICT and AI policies, which take precedence over this statement where they conflict
• − Removed: Legislative compliance statement removed
• − Removed: Safeguard dropped: acceptable use policy
• + Added: AI use-case register: not mentioned → planned
• + Added: Accountable official: not mentioned → designated
• + Added: New commitment (human oversight): All AI-generated outputs reviewed and verified by an appropriately authorised NIAA staff member
• + Added: New commitment (will not): NIAA does not use AI for decision-making
• + Added: New commitment (will): All staff will complete the APS Academy AI in Government Fundamentals course
• + Added: New commitment (will): Establish a risk assessment approach aligned to the DTA AI Impact Assessment
• + Added: New commitment (will): Maintain an internal AI Use Case Register
• + Added: New commitment (will): Notify the DTA of transparency compliance and report high-risk AI applications via the DTA Reporting Portal
• + Added: Safeguard added: human review of outputs
• + Added: Safeguard added: privacy or security controls
• + Added: Safeguard added: risk assessment
• + Added: Safeguard added: use case register
• ~ Changed: Policy version referenced: v1 → v2
• ~ Changed: Stated first-published date: 2025-02-19 → 2026-03-27

TEQSA rewrote its statement. It now names a Chief AI Officer and an AI Steering Group, discloses AI uses it is considering, and commits to human review of public-facing AI.

• Chief AI Officer appointed and AI Steering Group established (previously the Steering Committee was only planned)
• Considering Microsoft Copilot analytics, in the corporate and enabling domain and the compliance and fraud detection domain
• Commits not to use AI for decision-making, or where the public or providers interact with AI or are significantly affected by it without human review
• New effectiveness monitoring: training completion, productivity, insights and AI incidents

• − Removed: Commitment dropped (will not): Has not authorised AI use in service delivery, including regulatory function or decision-making processes
• − Removed: Commitment dropped (will): Statement will be reviewed annually or when significant change is made to AI approach
• − Removed: Strategic position on AI: planned → not mentioned
• + Added: Chief AI Officer: not mentioned → in place
• + Added: Commitment: no public-facing AI without a human added
• + Added: Intentions behind AI use added
• + Added: Legislative compliance statement added
• + Added: Monitoring and protection measures added
• + Added: New commitment (human oversight): Not considering AI use where public or providers interact with or are significantly impacted without human review
• + Added: New commitment (will not): Not considering adoption of AI for decision-making purposes
• + Added: New commitment (will): Any AI use will follow the requirements of the DTA's Policy for the responsible use of AI in government
• + Added: New commitment (will): Will monitor training completion, productivity impacts, AI insights and AI incidents, updating governance accordingly
• + Added: New commitment (will): Will only use AI in accordance with applicable legislation, frameworks, policies and best practice
• + Added: New commitment (will): Will only use AI where assessed and found to be safe, ethical and effective
• + Added: Policy compliance statement added
• + Added: Safeguard added: incident or concern reporting
• + Added: Safeguard added: risk assessment
• ~ Changed: Commitment kept (will): An AI Steering Committee will be established to develop an overarching agency approach to AI
• ~ Changed: Review cadence: annual and on change → not stated
• ~ Changed: Stated first-published date: 2025-03 → —
• ~ Changed: Stated last-updated date: 2025-03-25 → 2026-03-25

Adds a "last updated 25 March 2026" date and a matching note in the updates section; a page-feedback widget also drops out of the capture.

Names its AI Branch Manager, Ramsey Beydoun, as an accountable official in place of the Chief Technology Officer, and drops its commitment to update the statement at least every twelve months.

• Accountable official changed: CTO Andrew Morrison replaced by AI Branch Manager Ramsey Beydoun
• Removed: commitment to update the statement as its approach to AI changes and at least every twelve months
• Text still gives the CTO primary responsibility for policy areas, which no longer matches the named official

• − Removed: Commitment dropped (will): Statement updated as approach to AI changes, and at least every twelve months
• ~ Changed: Review cadence: annual and on change → not stated

Only the OAIC website's navigation menu and page-feedback widget changed; the AI transparency statement text is unchanged.

Replaces its own detailed statement with a short referral to the Department of Social Services' statement, handing the Chief AI Officer role to DSS and dropping its tool disclosures and safeguards.

• Department of Social Services to hold the SES Chief AI Officer role for the Commission
• Commission now relies on the DSS AI transparency statement, with a link to it
• Accountable official renamed Director - Corporate Support (EL2), replacing Director - Governance and Corporate Coordination
• New commitment to update the page when information changes

• − Removed: Commitment dropped (human oversight): AI supports operations but does not replace human judgment in areas affecting people's lives
• − Removed: Commitment dropped (human oversight): All AI outputs are checked by a human before they are applied or acted on
• − Removed: Commitment dropped (will): Develop training and support information relevant to the Commission
• − Removed: Commitment dropped (will): Ensure AI use is appropriately governed and risks are identified and managed
• − Removed: Commitment dropped (will): Ensure compliance with applicable legislation and regulations in any AI use
• − Removed: Commitment dropped (will): Make efforts to protect the public against negative impacts of AI
• − Removed: Commitment dropped (will): Monitor AI access and usage and the effectiveness of deployed AI systems
• − Removed: Commitment dropped (will): Review and update the statement annually or on significant change to approach to AI
• − Removed: Domain dropped: corporate and enabling
• − Removed: Domain dropped: policy and legal
• − Removed: Intentions behind AI use removed
• − Removed: Legislative compliance statement removed
• − Removed: Monitoring and protection measures removed
• − Removed: Named tool dropped: Microsoft 365 Copilot Chat
• − Removed: Safeguard dropped: human review of outputs
• − Removed: Safeguard dropped: privacy or security controls
• − Removed: Safeguard dropped: risk assessment
• − Removed: Safeguard dropped: staff training
• − Removed: Staff training: mandatory → not mentioned
• − Removed: Usage pattern dropped: analytics for insights
• − Removed: Usage pattern dropped: workplace productivity
• + Added: Chief AI Officer: not mentioned → in place
• + Added: New commitment (will): Accountable official will continue working with DTA and the Department on the Commission's responsible AI use
• + Added: New commitment (will): Inform the public by updating this webpage when information changes or updates are required
• ~ Changed: Review cadence: annual and on change → on change
• ~ Changed: Stated last-updated date: 2026-02-03 → 2026-03-19

Writes out "Chief AI Officer" in full in place of the abbreviation "CAIO". The appointment and the role's responsibilities are unchanged.

No change to AUSTRAC's statement; the page's feedback widget and "About us" navigation menu are no longer captured.

Rewrites its statement to name ChatGPT, Claude and Microsoft 365 Copilot, and to require stakeholders to declare AI use. Drops its commitment to review the statement at least every 12 months.

• Names tools in use: Microsoft 365 Copilot (enterprise), and public ChatGPT and Claude
• Internal AI policy updated March 2026 to incorporate the APS AI Plan 2025 and the staff guidance on public generative AI; mandatory staff training introduced
• Stakeholders, including funding applicants, must now declare their AI use
• Funding application forms have collected information on AI use since 2025, for research purposes only

• − Removed: Commitment dropped (will): Staff may use AI tools only if the agency has approved them
• − Removed: Commitment dropped (will): Statement updated as approach to AI changes, and at least every 12 months
• + Added: Named tool added: ChatGPT
• + Added: Named tool added: Claude
• + Added: Named tool added: Microsoft 365 Copilot
• + Added: New commitment (will): Continually refine AI policies to align with Government guidance, regulation, legislation and best practice
• + Added: New commitment (will): Contractors and consultants must follow security requirements and disclose and document their AI use
• + Added: New commitment (will): Stakeholders, including funding applicants, must declare any AI use
• ~ Changed: Public-facing AI use: with human review → none
• ~ Changed: Review cadence: annual and on change → not stated

Treasury has rewritten its statement. It names Copilot Chat and allows some public AI tools, commits to appointing a Chief AI Officer by July 2026, and moves its update date to February 2026.

• Commits to appointing a Chief AI Officer by July 2026 under the APS AI Plan
• Endorses Microsoft Copilot Chat alongside M365 Copilot; Copilot no longer limited to trial participants
• Permits some publicly available AI tools under the DTA's staff guidance on public generative AI
• States that AI makes no decisions and all AI use has human oversight

• + Added: Chief AI Officer: not mentioned → planned
• + Added: New commitment (human oversight): All AI use has human oversight; decisions are not made by AI
• + Added: New commitment (will): A Chief AI Officer will be appointed by July 2026 to lead AI transformation
• ~ Changed: Public-facing AI use: with human review → none
• ~ Changed: Stated last-updated date: 2025-02-28 → 2026-02-28

Tidies capitalisation, turns a bold label into a proper heading for the Australian AI Safety Institute section and trims one word; nothing of substance changes.

Rewrites its statement: Copilot is now in use rather than planned, a February 2026 staff working group is disclosed, and officer names and several earlier commitments are dropped.

• Microsoft Copilot now available to all staff and Microsoft 365 Copilot to approved roles (previously only a planned trial)
• New: staff working group set up in February 2026 to assess wider AI use, including submission analysis, proofing and document handling
• Names LexisNexis AI+ and Westlaw Edge as legal research tools under consideration, not currently used for inquiry work
• New: states AI is not used for law reform work (legal analysis, drafting or recommendations) or in ways the public directly interacts with

• − Removed: Commitment dropped (will): Staff must complete internal generative AI training before using Copilot
• + Added: Domain added: corporate and enabling
• + Added: Named tool added: Microsoft 365 Copilot
• + Added: New commitment (will not): Do not use AI for law reform activities, including legal analysis, drafting or recommendations
• + Added: New commitment (will not): Use AI only for approved, low-risk workplace productivity purposes
• + Added: New commitment (will): Update the statement if AI becomes public-facing or expands to new usage patterns or domains
• + Added: Strategic position on AI: not mentioned → in place
• + Added: Usage pattern added: workplace productivity
• ~ Changed: Commitment kept (will): Implement AI fundamentals training for all staff, regardless of role

Adds Service Delivery as an AI domain, adds meeting summarisation, recasts its governance refresh as a year's work, and says it is planning how to comply with policy v2.0.

• AI use now disclosed in the Service Delivery domain as well as Corporate & Enabling
• Meeting summarisation added to the workplace productivity uses
• New statement that it is planning how to meet policy v2.0's additional requirements within the required timeframes
• Governance refresh now described as work done over the past year rather than work currently under way

• + Added: Domain added: service delivery
• + Added: New commitment (will): Plan approach to comply with additional policy v2.0 requirements within required timeframes
• ~ Changed: Policy version referenced: unspecified → v2

Rewrites its statement to report an AI Policy approved in December 2025, discloses AI regulatory risk models, and names two AI accountable officials instead of the Chief Digital Officer.

• Accountable official changed from the Chief Digital Officer to two officials: the Chief Digital Information Officer and the Director of AI
• Discloses that AI analytics include regulatory risk models
• New fact: Commission AI Policy approved 10 December 2025
• New commitment to update the statement with its AI adoption strategy by 15 June 2026

• + Added: New commitment (will): Will update the transparency statement to reflect the AI strategic position by 15 June 2026
• + Added: Safeguard added: acceptable use policy
• ~ Changed: Stated last-updated date: 2025-02 → 2026-02

Rewrites its statement for policy v2.0: names a shared Chief AI Officer, drops mention of its website chatbot, says the public never interacts with its AI, and adds an email contact.

• Chief AI Officer role now shared by the CIO and the Regulatory Transformation Group Manager; the CIO remains the AI Accountable Official
• New commitment: no AI where the public may directly interact with it or be significantly affected by it
• Removed: disclosure of a website chatbot and language-conversion tools
• Classification widened from workplace productivity and analytics to include compliance and fraud detection, law enforcement and security, policy and legal, and corporate uses

• − Removed: Commitment dropped (will): AI governance framework will follow OAIC best practice guidance on commercial AI and personal information
• − Removed: Commitment dropped (will): All AI initiatives will align with the Protective Security Policy Framework and Information Security Manual
• − Removed: Commitment dropped (will): Consult employees about the potential use of AI in the workplace
• − Removed: Commitment dropped (will): Develop internal policies to guide AI use and provide training to build staff capability
• − Removed: Commitment dropped (will): Establish a new internal governance body to provide strategic direction and monitor AI initiatives
• − Removed: Commitment dropped (will): If personal information is used in AI, comply with Australian Privacy Principles and explain its use transparently
• − Removed: Domain dropped: service delivery
• − Removed: Named tool dropped: Microsoft 365 Copilot
• − Removed: Safeguard dropped: acceptable use policy
• − Removed: Safeguard dropped: governance body
• − Removed: Safeguard dropped: privacy or security controls
• − Removed: Safeguard dropped: staff training
• − Removed: Staff training: available → not mentioned
• + Added: Chief AI Officer: not mentioned → in place
• + Added: Domain added: compliance and fraud detection
• + Added: Domain added: law enforcement intelligence and security
• + Added: Domain added: policy and legal
• + Added: New commitment (will): Formalise strategic approach with AI Strategy 2026–27 and AI Governance Framework in March 2026
• + Added: Safeguard added: human review of outputs
• + Added: Strategic position on AI: not mentioned → in place
• ~ Changed: Policy version referenced: unspecified → v2
• ~ Changed: Public-facing AI use: unclear → none

Finance rewrites its statement: it names a Chief AI Officer, describes its lead role under the APS AI Plan and GovAI Chat, sets data limits by tool, and replaces its two co-chairing officials.

• Chief AI Officer appointed (FAS, AI Delivery and Enablement Division); co-chairs the AI Governance Committee and co-authorises the statement with the Accountable Official
• Removed: second accountable official (FAS, Risk, Claims and Regulatory Reform Division) and the list of committee member divisions
• New disclosure: Finance develops and runs GovAI Chat and leads APS AI Plan delivery through its AIDE function
• Microsoft 365 Copilot may be used with information up to PROTECTED; public AI tools are limited to OFFICIAL, with no personal or sensitive data

• − Removed: Commitment dropped (human oversight): Limit AI to low-risk use cases: no direct public interaction or significant impact without human intervention
• − Removed: Commitment dropped (will): Reassess use cases on notable change, life-cycle progression, or identified AI risk or harm
• − Removed: Commitment: no public-facing AI without a human removed
• − Removed: Named tool dropped: ChatGPT
• − Removed: Named tool dropped: Claude
• − Removed: Named tool dropped: Gemini
• + Added: AI use-case register: not mentioned → in place
• + Added: Chief AI Officer: not mentioned → in place
• + Added: Named tool added: GovAI Chat
• + Added: New commitment (will not): Do not enter personal, sensitive or classified information into public generative AI tools
• + Added: New commitment (will not): Won't approve AI solutions or uses still carrying medium or high risk after safeguards
• + Added: New commitment (will): Committed to full implementation of the Policy and ongoing compliance
• + Added: New commitment (will): Defer deployment where residual risk is uncertain until risks are appropriately mitigated
• + Added: New commitment (will): Uses beyond central approvals require an AI impact assessment and registration
• + Added: Safeguard added: incident or concern reporting
• + Added: Safeguard added: use case register
• ~ Changed: Public-facing AI use: with human review → not addressed

Drops its statement that AI isn't used in service delivery or decision-making. It now discloses AI supporting service delivery and human decisions, and adds governance for AI used by third-party providers.

• Removed: statement that DEWR does not use AI in service delivery, program administration, compliance, auditing or decision-making
• New usage pattern: AI supports human decision-making and administrative action (states AI is not used for automated decision-making)
• New domain: service delivery, including services delivered by contracted third-party providers
• New Third-Party AI Assessment Framework and AI in Procurement Principles for overseeing providers' use of AI

• + Added: Domain added: service delivery
• + Added: New commitment (will not): AI is not used for automated decision making; it only supports human decision makers
• + Added: New commitment (will): DEWR remains accountable for third-party AI use in contracted services, overseen via Third-Party AI Assessment Framework
• + Added: Safeguard added: risk assessment
• + Added: Usage pattern added: decision making and administrative action
• ~ Changed: Public-facing AI use: none → with human review

The ATO discloses AI use in law enforcement and intelligence and in image processing, commits to policy v2.0 and a standalone AI policy, and drops its claim that it doesn't use AI for administrative action.

• Removed: the note that the ATO does not use AI for administrative action
• New domain disclosed: law enforcement, intelligence and security (AI-enabled analysis of data for intelligence gathering)
• New usage pattern disclosed: image processing
• New uses: alerting taxpayers to lodgments suspected of identity crime; improving the readability of public advice content

• − Removed: Commitment dropped (will not): We do not use AI for administrative action
• + Added: Domain added: law enforcement intelligence and security
• + Added: New commitment (will not): AI use limited to approved tools used for approved use cases under our AI policy
• + Added: New commitment (will): Committed to implementing version 2.0 policy requirements from 15 December 2025
• + Added: New commitment (will): Will soon release a standalone AI policy setting staff obligations to support version 2.0
• + Added: Safeguard added: acceptable use policy
• + Added: Usage pattern added: image processing
• ~ Changed: Policy version referenced: unspecified → v2
• ~ Changed: Stated first-published date: 2025-02-28 → —

Moves from exploring AI to disclosing actual use of Microsoft 365 Copilot by all staff, and limits mandatory AI training to staff holding Copilot licences.

• Discloses current generative AI use: all staff have Microsoft 365 Copilot Chat, and a limited number hold full Copilot licences
• Training now mandatory only for licensed Copilot users and 'strongly encouraged' for others, replacing the promise of mandatory training for all staff before any AI use
• Classifies its use under the DTA domains of policy and legal, scientific, and corporate and enabling
• Adds staff principles: protect privacy, critically assess outputs, and own their advice and decisions

• − Removed: Commitment dropped (will): Ethical and responsible AI training mandatory for all staff before using any AI tools
• + Added: Domain added: corporate and enabling
• + Added: Domain added: policy and legal
• + Added: Domain added: scientific
• + Added: Named tool added: Microsoft 365 Copilot
• + Added: New commitment (will): Ethical and responsible AI training mandatory for all staff with a Microsoft 365 Copilot licence
• + Added: New commitment (will): Update statement on completing internal review against Policy v2.0
• + Added: Usage pattern added: analytics for insights
• + Added: Usage pattern added: workplace productivity
• ~ Changed: Stated last-updated date: — → 2026-02-26

ASIC reviewed its statement for the December 2025 policy: it names its accountable official as Chief AI Officer, adds a use case register and impact assessment, and plans mandatory staff training.

• Senior Executive Leader, Data, Analytics and AI named as chief AI officer under the APS AI Plan 2025, as well as accountable official
• Internal AI use case register added, capturing the Standard's minimum fields, with a named owner for each use case
• AI use case assessment covers all provisions of the DTA's AI impact assessment tool; high-risk use cases reported to the DTA
• AI board role expanded to approving use cases, risk mitigation and ongoing monitoring

• + Added: AI use-case register: not mentioned → in place
• + Added: Chief AI Officer: not mentioned → in place
• + Added: New commitment (will): Will implement APS Academy AI in government fundamentals as mandatory training for all ASIC staff
• + Added: New commitment (will): Will report any high-risk AI use cases to the DTA
• + Added: Safeguard added: incident or concern reporting
• + Added: Safeguard added: use case register
• + Added: Strategic position on AI: not mentioned → planned
• ~ Changed: Policy version referenced: unspecified → v2
• ~ Changed: Stated last-updated date: 2025-02-28 → 2026-02-27

Refreshes its statement for February 2026. It discloses new AI uses including document drafting and task automation, adopts the government policy's definition of AI, and names its Corporate Governance Board as ultimate overseer.

• New uses disclosed: automating routine tasks, editing drafts, generating drafts (except documents subject to administrative review), proactive information seeking for analysis
• Corporate Governance Board named as having ultimate oversight of AI adoption; working group now supports it
• New commitment to update the statement if AI is used in any public interaction
• New claim that AI systems do not make decisions or take actions on their own and that AI use complies with all relevant law

• ~ Changed: Stated first-published date: 2025-02 → 2026-02

The Department of Veterans' Affairs statement didn't change; the scraper just stopped picking up the page's "On this page" menu and the "Was this page useful?" feedback form.

The department discloses that it is responsible for the new Australian AI Safety Institute. It also cites version 2.0 of the policy and moves its updated date to February 2026.

• Added: the department is responsible for the Australian AI Safety Institute, which is being set up under the National AI Plan to monitor and test frontier AI
• Now cites version 2.0 of the DTA policy and drops the 1 September 2024 commencement date
• AI use described as 'a number of AI tools' instead of 'a small number'
• Date updated changed from 26 February 2025 to 26 February 2026

• ~ Changed: Policy version referenced: v1 → v2
• ~ Changed: Stated last-updated date: 2025-02-26 → 2026-02-26

The final sentence, on annual review, is missing from this capture and its "Review and Updates" heading is left empty, which points to a partial capture rather than an edit.

Rewrites its statement: names Copilot tools and an Accountable Official, promises a Chief AI Officer in 2026, adds mandatory staff training, and switches to annual review.

• Discloses Microsoft 365 Copilot and Copilot Chat as the endorsed tools; some public AI tools permitted under DTA guidance
• Designates the Manager, Data & Digital Solutions as Accountable Official
• Says a Chief AI Officer will be appointed during 2026
• Adds mandatory APSC AI Fundamentals course and internal training before staff can use enterprise AI tools

• − Removed: Safeguard dropped: governance body
• + Added: Chief AI Officer: not mentioned → planned
• + Added: Legislative compliance statement added
• + Added: Named tool added: Microsoft Copilot Chat
• + Added: New commitment (will): Appoint a Chief AI Officer during 2026
• + Added: New commitment (will): Review and update statement annually or whenever significant changes occur
• + Added: New commitment (will): Staff must complete internal AI training before using Enterprise AI tools
• + Added: Public contact added
• + Added: Safeguard added: human review of outputs
• + Added: Safeguard added: privacy or security controls
• + Added: Safeguard added: staff training
• + Added: Staff training: not mentioned → mandatory
• ~ Changed: Commitment kept (will): Review statement by 28 February 2026, on significant change, or when accuracy materially affected
• ~ Changed: Public-facing AI use: with human review → none
• ~ Changed: Stated last-updated date: 2025-02 → 2026-02-28

The Ombudsman's office discloses two new AI uses, AI coding assistants and Azure text-to-speech for internal training materials, and records a December 2025 update in its version history.

• New use disclosed: AI coding assistants for developing, configuring, maintaining and testing commercial software, limited to digital specialists, with all code tested before deployment
• New use disclosed: Azure AI Text to Speech for internal training materials from human-written scripts
• Security monitoring use now explicitly states there is no direct public access
• Pilots of Microsoft Copilot and Azure AI no longer described as 'recently' completed

• + Added: Named tool added: Azure AI Text to Speech
• + Added: New commitment (will): AI coding assistants used only for approved use cases; all AI-assisted code tested for accuracy before implementation
• + Added: Safeguard added: testing or evaluation
• + Added: Usage pattern added: workplace productivity
• ~ Changed: Stated first-published date: 2025-02-26 → 2025-02-27
• ~ Changed: Stated last-updated date: — → 2025-12-20

Changes the contact email for its AI Accountable Official and retitles the statement heading; the other changes are site navigation and a page date stamp.

• Contact email for the AI Accountable Official changed from enquires@nhfb.gov.au to nhfb.enquiries@nhfb.gov.au
• Page now shows a date stamp of 28 February 2025

Replaces its "no AI in use" statement with a full rewrite that discloses Copilot Chat, an AI-enabled contact centre and planned LLM document annotation, and changes the enquiries contact.

• Removed: statement that the Commission does not use AI internally or externally
• Discloses Microsoft 365 Copilot Chat for workplace productivity
• Discloses cloud contact platform using AI for text-to-speech, transcription and customer sentiment analysis
• Discloses planned LLM annotation of complaints, reportable incidents and audit reports, scheduled for February 2026

• + Added: Domain added: compliance and fraud detection
• + Added: Domain added: corporate and enabling
• + Added: Domain added: policy and legal
• + Added: Domain added: scientific
• + Added: Domain added: service delivery
• + Added: Intentions behind AI use added
• + Added: Legislative compliance statement added
• + Added: Monitoring and protection measures added
• + Added: Named tool added: Microsoft 365 Copilot Chat
• + Added: New commitment (human oversight): Always maintain appropriate human oversight; staff review and finalise Copilot content
• + Added: New commitment (will not): AI is not used to automate recruitment, selection or employment-related decision making
• + Added: New commitment (will not): Browser-based AI not used to make, automate or determine regulatory or compliance decisions
• + Added: New commitment (will not): No regulatory decisions are made by AI; models do not recommend regulatory actions
• + Added: New commitment (will): Internal AI policies will align with DTA Policy and apply to all employees and contractors
• + Added: New commitment (will): Staff must complete internal AI training before using AI tools
• + Added: New commitment (will): Will comply with mandatory requirements of the DTA Policy for the Responsible Use of AI (v2.0)
• + Added: New commitment (will): Will comply with relevant legislation, regulations, frameworks and best practice standards
• + Added: New commitment (will): Will continue to train and develop staff to use AI safely, responsibly and ethically
• + Added: New commitment (will): Will continue working closely with the DTA and follow applicable guidelines
• + Added: New commitment (will): Will monitor AI tools through governance and act if benefits unrealised or unintended consequences arise
• + Added: Policy compliance statement added
• + Added: Safeguard added: acceptable use policy
• + Added: Safeguard added: human review of outputs
• + Added: Safeguard added: privacy or security controls
• + Added: Safeguard added: risk assessment
• + Added: Safeguard added: staff training
• + Added: Staff training: not mentioned → mandatory
• + Added: Usage pattern added: analytics for insights
• + Added: Usage pattern added: decision making and administrative action
• + Added: Usage pattern added: workplace productivity
• ~ Changed: Commitment kept (will): Will update this statement in accordance with the Policy when AI capability is implemented
• ~ Changed: Policy version referenced: v1 → v2
• ~ Changed: Public-facing AI use: none → with human review
• ~ Changed: Stated first-published date: — → 2026-02
• ~ Changed: Stated last-updated date: 2025-02 → 2026-02

Rewrites its statement to disclose that it now uses generative AI tools, including GovAI, Microsoft 365 Copilot and GitHub Copilot. It adds human review of all outputs, rules out AI decision-making and adds a dedicated AI contact email.

• Removed: statement that the NBA does not use AI in any services
• New AI uses disclosed: GovAI Multi-Model (Jan 2026), Microsoft 365 Copilot (Feb 2026), Copilot via GovTEAMS (Jan 2026), GitHub Copilot (Feb 2026)
• Usage classified as analytics for insights and workplace productivity, across service delivery, policy and legal, scientific and corporate domains
• New commitment: a human assesses all AI outputs and AI is not used for decision-making

• − Removed: Commitment dropped (will): Statement will be reviewed annually or on any significant change to our approach to AI
• + Added: Legislative compliance statement added
• + Added: Monitoring and protection measures added
• + Added: Named tool added: GitHub Copilot
• + Added: Named tool added: GovAI
• + Added: Named tool added: Microsoft Copilot
• + Added: New commitment (human oversight): A human will assess all AI outputs to ensure accuracy
• + Added: New commitment (will not): Will not use AI for decision making purposes
• + Added: New commitment (will): AI adoption will align with the DTA's Policy, advice and guidance on safe, responsible AI
• + Added: New commitment (will): Will develop internal AI policies and guidance (governance, acceptable use, records, risk, training) applying to all staff
• + Added: New commitment (will): Will update this transparency statement if our approach changes
• + Added: New commitment (will): Will use AI in accordance with applicable legislation, regulations, frameworks, policies and best practice
• + Added: Policy compliance statement added
• + Added: Safeguard added: human review of outputs
• + Added: Usage pattern added: analytics for insights
• + Added: Usage pattern added: workplace productivity
• ~ Changed: Review cadence: annual and on change → on change
• ~ Changed: Stated first-published date: 2025-01 → —

Names a Chief AI Officer, gives the accountable official role to a different officer, discloses new AI uses such as transcription and description extraction, and adds concrete safeguards like mandatory training.

• Assistant Director-General Technology named Chief Artificial Intelligence Officer
• Accountable official changed from the Assistant Director-General Technology (as CIO) to the Chief Information and Data Governance Officer
• New AI uses disclosed: transcription and content extraction, description extraction and enhancement, and trialling AI to improve archival processes
• Safeguards added: AI tools and users registered, tools assessed for cyber security, mandatory staff AI training, all AI use declared, and risk and ethics assessments for business use cases

• − Removed: Commitment dropped (will): Will continue to update this transparency statement to reflect implementation and adoption of AI
• + Added: AI use-case register: not mentioned → in place
• + Added: Chief AI Officer: not mentioned → in place
• + Added: Safeguard added: privacy or security controls
• + Added: Safeguard added: staff training
• + Added: Safeguard added: use case register
• + Added: Staff training: not mentioned → mandatory
• ~ Changed: Stated last-updated date: 2025-07-02 → 2026-01-07

Nothing Geoscience Australia wrote changed; the breadcrumb navigation now shows an extra "Geoscience Australia" link captured from the page chrome.

DVA rewrites its statement: it says its AI Policy is now finalised, discloses new AI uses and trials, is formalising a Chief AI Officer, and drops its annual review commitment.

• New use disclosed: semantic text analysis of MyService claim responses, with staff keeping responsibility for claim decisions
• New use disclosed: Microsoft 365 Copilot Chat in production with DVA-specific safeguards; licensed M365 Copilot now in trial
• New trial planned: a voluntary, consent-based proof-of-concept Claims Document Reader built on GovAI, with decisions left to trained delegates
• AI Policy now described as in place (previously 'finalising')

• − Removed: Commitment dropped (human oversight): All CLIKChat information is checked by staff, with human judgement always in control
• − Removed: Commitment dropped (will): Will monitor usage to assure ongoing suitability of AI technologies
• − Removed: Commitment dropped (will): Will review and update the statement annually or when significant changes occur
• + Added: Chief AI Officer: not mentioned → planned
• + Added: New commitment (human oversight): Claims Document Reader will not decide claims; trained DVA delegates continue to make claims decisions
• + Added: New commitment (human oversight): GovAI claim-processing trial uses no veteran personal data and always has human oversight
• + Added: New commitment (human oversight): Staff always retain responsibility for analysis and claim decisions when Semantic Text Analysis is used
• + Added: New commitment (will): Proposed Claims Document Reader trial will be voluntary with a clear consent model for participants
• + Added: New commitment (will): Will formalise designation of a Chief AI Officer
• + Added: Safeguard added: acceptable use policy
• + Added: Safeguard added: staff training
• + Added: Staff training: not mentioned → mandatory
• + Added: Usage pattern added: decision making and administrative action
• ~ Changed: Commitment kept (will): Will finalise an AI Policy covering principles, risk management and approval considerations
• ~ Changed: Policy version referenced: v1 → unspecified
• ~ Changed: Review cadence: annual and on change → not stated
• ~ Changed: Stated first-published date: 2025-02 → —
• ~ Changed: Stated last-updated date: 2025-10-09 → 2026-02-09

Discloses a new AI-assisted evaluation model for Digital Marketplace Panel 2 decisions, adds Service Delivery as an AI domain, and says AI training is now in place.

• New use of AI: structured AI-assisted evaluation of Digital Marketplace Panel 2 (DMP2) applications from February 2026, making recommendations to a human decision maker
• Adds a new section classifying use cases: assisting administrative decision-making, and workplace productivity through Microsoft 365 Copilot
• AI domains broadened from Corporate and Enabling to also cover Service Delivery
• AI fundamentals training for all staff described as implemented, no longer planned

• + Added: Domain added: service delivery
• + Added: New commitment (human oversight): AI assessment of DMP2 applications makes recommendations to a human decision maker
• + Added: Safeguard added: human review of outputs
• + Added: Usage pattern added: decision making and administrative action
• ~ Changed: Commitment kept (will): Will implement AI fundamentals training for all staff, regardless of role
• ~ Changed: Public-facing AI use: none → with human review
• ~ Changed: Stated last-updated date: 2024-12-04 → 2025-12-01

Rewrites the statement: DSS moves from exploratory trials to approved AI use cases, discloses new tools, names a Chief AI Officer and a governance committee, and adds a compliance table.

• New tools disclosed: Microsoft 365 Copilot Chat (internal) and Otter AI transcription for selected staff
• Chief AI Officer appointed; AI Governance Committee chaired by the COO reviews and approves use cases
• Classifies use under the DTA scheme and rules out decision-making, image processing, compliance/fraud, scientific and law enforcement uses
• Commits that no use involves direct interaction with the public or significantly affects individuals, and that a human checks every output

• − Removed: Commitment dropped (will): Will make efforts to protect the public against negative impacts
• − Removed: Commitment dropped (will): Will review and update the statement annually or on significant change to AI approach
• + Added: AI use-case register: not mentioned → planned
• + Added: Chief AI Officer: not mentioned → in place
• + Added: Domain added: service delivery
• + Added: Named tool added: GovAI
• + Added: Named tool added: Otter.ai
• + Added: New commitment (human oversight): All AI outputs are checked by a human before being applied or acted on
• + Added: New commitment (will not): Technological controls restrict access to publicly available AI tools from the DSS ICT environment
• + Added: New commitment (will): Every AI use case will have approved KPIs to track outcomes and measure effectiveness
• + Added: New commitment (will): Framework will require all AI use cases to be assessed in accordance with the DTA policy
• + Added: New commitment (will): Will document an AI strategic position by 1 June 2026
• + Added: New commitment (will): Will implement AI governance framework, use case owners and internal use case register by 1 December 2026
• + Added: Public contact added
• + Added: Safeguard added: acceptable use policy
• + Added: Safeguard added: governance body
• + Added: Safeguard added: human review of outputs
• + Added: Safeguard added: privacy or security controls
• + Added: Staff training: available → mandatory
• + Added: Strategic position on AI: not mentioned → planned
• ~ Changed: Commitment kept (will): Will roll out AI training throughout the year to build workforce AI capability
• ~ Changed: Policy version referenced: unspecified → v2
• ~ Changed: Public-facing AI use: not addressed → none
• ~ Changed: Review cadence: annual and on change → not stated
• ~ Changed: Stated last-updated date: 2025-08-28 → 2026-01-22

Rewrites its statement for policy v2.0. It now discloses Copilot use and training requirements, an AI subcommittee, an assurance framework and a use case register, drops the no-public-interaction claim and names a new AI team contact.

• Discloses the 2024 Microsoft 365 Copilot trial, Copilot Chat for all staff and a phased rollout of licensed Copilot
• Staff must complete AI fundamentals training before using Copilot
• New AI Subcommittee (AISc) advising the Digital Committee
• Discloses an AI assurance framework and an internal AI use case register that limits tools to approved uses

• − Removed: Commitment dropped (human oversight): A human will assess all AI outputs to ensure accuracy
• − Removed: Commitment dropped (will not): Will not use AI for decision making purposes
• − Removed: Commitment dropped (will): Ensure policies meet proposed mandatory guardrails for AI in high-risk settings
• − Removed: Commitment dropped (will): Update this transparency statement if our approach changes
• − Removed: Safeguard dropped: human review of outputs
• + Added: AI use-case register: not mentioned → in place
• + Added: Domain added: policy and legal
• + Added: Named tool added: Microsoft 365 Copilot Chat
• + Added: New commitment (human oversight): Staff will explain, justify and take ownership of advice and decisions informed by AI
• + Added: New commitment (will not): Do not use AI to automate decisions; human officials remain fully accountable
• + Added: New commitment (will not): Restrict AI tools to approved use cases in the use case register
• + Added: New commitment (will not): Staff cannot use sensitive or personal information without assurance and governance approval
• + Added: New commitment (will): Continuously review AI use to protect the public and align with whole-of-government guidance
• + Added: New commitment (will): Staff must complete AI fundamentals training and acknowledge responsible use before accessing Copilot
• + Added: Safeguard added: acceptable use policy
• + Added: Safeguard added: audit or assurance
• + Added: Safeguard added: governance body
• + Added: Safeguard added: incident or concern reporting
• + Added: Safeguard added: privacy or security controls
• + Added: Safeguard added: risk assessment
• + Added: Safeguard added: staff training
• + Added: Safeguard added: use case register
• + Added: Staff training: not mentioned → mandatory
• + Added: Usage pattern added: decision making and administrative action
• + Added: Usage pattern added: image processing
• ~ Changed: Policy version referenced: v1 → v2
• ~ Changed: Public-facing AI use: none → not addressed
• ~ Changed: Stated last-updated date: 2025-05-20 → 2026-02-16

Rewrites the statement in first person, says its AI ethics framework is now published, aligns its register with DTA's newer standards and drops its promise to report high-risk uses to the DTA.

• AI ethics and governance framework now described as developed and published, not a future commitment
• Register and risk assessment now aligned with DTA technical standard, AI Impact Assessment Tool and AI procurement guidance
• Removed: alignment with the Commonwealth AI Assurance framework pilot and its risk matrix
• Removed: commitment to report high-risk use cases to the DTA

• − Removed: Commitment dropped (will): Report any high-risk use cases surfaced through the register to the DTA in line with the Policy
• + Added: New commitment (will): Finalise AI Risk Appetite Guidelines to support AI risk mitigation
• ~ Changed: Commitment kept (will): Develop and publish an AI ethics and governance framework aligned with whole-of-government policy
• ~ Changed: Stated last-updated date: 2025-02 → 2026-02

DAFF updates its statement for policy v2.0: it moves the Accountable Official role to its Chief Information and Security Officer and names a Chief AI Officer. It also adds compliance and fraud detection uses and drops mention of the AI Assurance Framework pilot.

• Accountable Official moved from the Chief Data Officer to the Chief Information and Security Officer
• Chief Digital and Data Officer named as Chief AI Officer under the APS AI Plan
• New usage domain disclosed: compliance and fraud detection
• AI use cases now go through an assurance process before being tracked in the internal register

• + Added: Chief AI Officer: not mentioned → in place
• + Added: Domain added: compliance and fraud detection
• + Added: New commitment (will): Staff are required to report their use of AI to the Digital Business Division for the internal register
• ~ Changed: Policy version referenced: unspecified → v2
• ~ Changed: Stated last-updated date: 2025-07-25 → 2026-02-16

Rewrites the statement: it no longer says it doesn't use AI, discloses Copilot and compliance uses, names a Chief AI Officer and TSCC oversight, and adds explicit prohibitions.

• Removed: statement that the agency doesn't currently use AI in any services or advice
• Discloses staff use of enterprise AI, including Microsoft 365 Copilot, for drafting, research, coding, minutes and search
• Discloses AI-assisted image analysis and SRES compliance and fraud detection under development, subject to human review
• Chief AI Officer appointed in July 2025 and added as an accountable official

• − Removed: Commitment dropped (will): Provide regular updates on AI initiatives and their outcomes on our website
• + Added: AI use-case register: not mentioned → planned
• + Added: Chief AI Officer: not mentioned → in place
• + Added: Commitment: no public-facing AI without a human added
• + Added: Domain added: compliance and fraud detection
• + Added: Named tool added: Microsoft 365 Copilot
• + Added: New commitment (will not): No public-facing AI without user disclosure, human-in-the-loop escalation path, accuracy safeguards and accessibility consideration
• + Added: New commitment (will not): Won't automate compliance, enforcement or eligibility decisions without human review and approval
• + Added: New commitment (will not): Won't generate or issue regulatory notices or legal instruments without human validation and sign-off
• + Added: New commitment (will not): Won't train or prompt AI with sensitive or personal information unless expressly assessed and approved
• + Added: New commitment (will not): Won't use generative AI for forecasting or analytical outputs without human validation, provenance controls and audit capture
• + Added: New commitment (will): Complete a risk-based AI suitability assessment before any AI development or deployment
• + Added: New commitment (will): Maintain an AI Use Case Register, updated when use cases are approved or materially change
• + Added: New commitment (will): Pause or discontinue AI use if risk thresholds, accuracy metrics or safeguards are not met
• + Added: Safeguard added: acceptable use policy
• + Added: Safeguard added: privacy or security controls
• + Added: Safeguard added: risk assessment
• + Added: Safeguard added: staff training
• + Added: Safeguard added: testing or evaluation
• + Added: Safeguard added: use case register
• + Added: Staff training: not mentioned → mandatory
• + Added: Usage pattern added: analytics for insights
• + Added: Usage pattern added: image processing
• + Added: Usage pattern added: workplace productivity
• ~ Changed: Public-facing AI use: none → with human review

Rewrites its statement: it now discloses using AI for drafting, research, analysis and minutes, where it previously said it used none, and it has dropped the Chief Security Officer title from its accountable official.

• Now discloses AI use (it previously said it had adopted no AI applications): drafting and editing, policy research, analysing information for themes, and producing minutes and transcripts
• Classifies its AI use under the DTA classification system as scientific, policy, corporate and enabling domains
• New commitment: AI is used with human oversight, the Authority does not rely on AI alone to make decisions, and staff must confirm that AI-assisted content is accurate
• States it has no AI systems that interact directly with the public or make decisions about individuals without human oversight

• − Removed: Commitment dropped (will): Will explore staff AI use with DCCEEW as ICT provider, within ICT system constraints
• − Removed: Commitment dropped (will): Will regularly review policy compliance through existing governance and risk management processes
• + Added: Domain added: corporate and enabling
• + Added: Domain added: policy and legal
• + Added: Domain added: scientific
• + Added: Legislative compliance statement added
• + Added: Monitoring and protection measures added
• + Added: New commitment (human oversight): Staff remain responsible for confirming AI-assisted content is accurate and appropriate
• + Added: New commitment (will not): Will not rely on AI alone to make decisions
• + Added: New commitment (will): AI use must comply with the Authority's information, privacy and security requirements
• + Added: New commitment (will): Staff must comply with internal AI policies aligned with government and DCCEEW policies
• + Added: Safeguard added: acceptable use policy
• + Added: Safeguard added: human review of outputs
• + Added: Safeguard added: privacy or security controls
• + Added: Usage pattern added: analytics for insights
• + Added: Usage pattern added: workplace productivity
• ~ Changed: Policy version referenced: v1 → unspecified
• ~ Changed: Stated first-published date: 2025-02 → —
• ~ Changed: Stated last-updated date: — → 2026-02

AUSTRAC rewrites its statement. It now discloses current AI-enabled analytics for detecting financial crime and new domains, reports trials of enterprise generative AI, adds a compliance table and makes AI training mandatory.

• Analytics for Insights is now a current usage pattern; the statement previously said its rules-based and machine learning methods did not meet the definition of AI
• Removed: statement that its statistical and rules-based modelling falls outside the AI definition
• Service Delivery and Law Enforcement, Intelligence and Security are now listed as current AI domains rather than planned ones
• Discloses trials of enterprise generative AI for internal productivity and service delivery

• − Removed: Commitment dropped (will): Apply proportionate and targeted risk mitigations for AI use cases
• − Removed: Commitment dropped (will): Evaluate AI pilots in alignment with the Policy for Responsible Use of AI and internal AUSTRAC AI Policy
• + Added: Domain added: compliance and fraud detection
• + Added: Domain added: law enforcement intelligence and security
• + Added: Domain added: policy and legal
• + Added: Domain added: service delivery
• + Added: New commitment (will): All AI activities will be subject to appropriate privacy assurance to comply with legislation and policies
• + Added: New commitment (will): Committed to implementing version 2.0 policy requirements from 15 December 2025
• + Added: Safeguard added: governance body
• + Added: Staff training: available → mandatory

Makes the Chief Information Officer its accountable official instead of the Chief Operating Officer, and reports that its mandatory safe-use-of-AI training is now developed.

• Accountable official changed from Chief Operating Officer to Chief Information Officer
• Safe-use-of-AI training now described as developed, not in development
• Training now said to be mandatory for staff, contractors and consultants; tribunal members are no longer listed among those who must take it

• + Added: Safeguard added: staff training
• + Added: Staff training: available → mandatory
• ~ Changed: Commitment kept (will): Developing mandatory safe use of AI training for members, staff, contractors and consultants

Changes its review of the statement from twice a year to once a year, which pushes the next scheduled review from 20 August 2025 to 13 March 2026.

• Review cadence changed from twice-annually to annually
• Next scheduled review moved from 20 August 2025 to 13 March 2026

• ~ Changed: Review cadence: other → annual and on change

Names its Chief Information Officer as Chief AI Officer, discloses Microsoft 365 Copilot Chat for all staff, and says key election operations stay manual.

• Chief Information Officer appointed Chief AI Officer under the APS AI Plan
• New disclosed use: Microsoft 365 Copilot Chat enabled for all staff
• New assurance: Copilot doesn't replace staff; it runs within Microsoft 365 enterprise security and isn't used to train AI models
• New statement: voting and counting are done manually, in line with electoral law

• + Added: Chief AI Officer: not mentioned → in place
• ~ Changed: Stated last-updated date: 2025-09-25 → 2026-02-09

Adds commitments on AI embedded in commercial software and on human oversight, but drops its pledge to update the statement at least every twelve months.

• Removed: commitment to update the statement if its approach to AI changes and at least every twelve months
• New commitment to monitor authorised software for embedded AI features
• New commitment to educate and guide staff on spotting and using embedded AI features responsibly
• Staff are encouraged to report AI features they find and must keep human oversight and quality assurance of outputs

• − Removed: Commitment dropped (will): Update the statement if the approach to AI changes, and at least every twelve months
• + Added: Monitoring and protection measures added
• + Added: New commitment (human oversight): Ensure staff maintain appropriate human oversight and quality assurance of all outputs
• + Added: New commitment (will): Encourage staff to report identified AI features to support awareness and transparency
• + Added: New commitment (will): Ensure ongoing compliance by embedding a culture balancing AI risk management and innovation
• + Added: New commitment (will): Monitor and evaluate authorised software on an ongoing basis to identify embedded AI functionality
• + Added: New commitment (will): Provide staff education and guidance on identifying and responsibly using AI features
• + Added: Safeguard added: human review of outputs
• + Added: Safeguard added: incident or concern reporting
• + Added: Safeguard added: staff training
• + Added: Staff training: not mentioned → available
• ~ Changed: Review cadence: annual and on change → not stated
• ~ Changed: Stated last-updated date: 2025-02-25 → 2025-11-20

DFAT has rewritten its statement to disclose its AI usage categories and domains. It also names the Chief Information Officer as responsible, adds use-case owners and human oversight, and commits to annual review.

• Discloses AI usage categories: analytics for insights, workplace productivity, image processing
• Discloses domains: service delivery, compliance and fraud detection, policy and legal, corporate and enabling
• Names the Chief Information Officer as responsible for compliance with whole-of-government AI policy
• Adds an Accountable Use Case Owner for each AI use case, plus risk assessment before implementation and regular review

• − Removed: Named tool dropped: Microsoft 365 Copilot
• − Removed: Safeguard dropped: acceptable use policy
• + Added: Domain added: compliance and fraud detection
• + Added: Domain added: policy and legal
• + Added: Domain added: service delivery
• + Added: New commitment (human oversight): AI used only while always maintaining human oversight ('Human-in-the-Loop')
• + Added: New commitment (will): Transparency statement will be updated at least every twelve months
• + Added: Safeguard added: human review of outputs
• + Added: Safeguard added: risk assessment
• + Added: Usage pattern added: analytics for insights
• + Added: Usage pattern added: image processing
• ~ Changed: Public-facing AI use: none → not addressed
• ~ Changed: Review cadence: on change → annual and on change
• ~ Changed: Stated last-updated date: 2025-02-06 → 2025-12-03

Moves its AI Accountable Official role from the General Manager Digital and Technology to the Head of Digital Products and Platforms.

• Accountable Official changed from General Manager Digital and Technology to Head of Digital Products and Platforms

Changes its accountable official under the AI policy from the Chief Information Officer to the Chief Operating Officer.

• Accountable official changed from Chief Information Officer to Chief Operating Officer
• COO now engages with the Senior Management Committee and Principal Registrar/CEO on AI approvals

Makes AI training mandatory before staff get a licence, requires staff to report concerns, and now says outright that it uses AI rather than that it may.

• AI training now mandatory for all staff before a licence is granted (previously staff only had access to it)
• Staff now obliged to report AI concerns to the IMT team (previously only encouraged)
• AI use stated as current ('We utilise'), no longer conditional ('We may utilise')
• Evaluation criteria for AI services now named as including privacy and security

• + Added: New commitment (will): All staff must complete training on appropriate AI use before a licence is granted
• + Added: Staff training: available → mandatory
• ~ Changed: Stated last-updated date: 2025-06-23 → 2025-10-29

Screen Australia's statement is unchanged; only the rendering of its Cloudflare-obscured enquiry email address differs between captures.

Nothing of substance changed: Cloudflare's email-obfuscation link on the AI contact address rendered differently, and a share-by-email widget dropped out of the capture.

The only change is Cloudflare re-encoding the obfuscated enquiries email link. The contact address and the rest of the statement are the same.

No change to the statement's text; only the email addresses that Cloudflare obfuscates were re-encoded with new keys, in the AI contact link and the share widget.

TEQSA's statement is unchanged; the only difference is that breadcrumb and sidebar navigation have dropped out of the capture.

Nothing Screen Australia wrote changed: the site's breadcrumb and sidebar navigation dropped out of the capture, and the obfuscated contact email link got a new encoding.

No change to the statement's content. The only difference is that the site's breadcrumb navigation no longer shows up in the capture.

No change to the Productivity Commission's statement. Only the breadcrumb navigation at the top of the page stopped being captured.

The breadcrumb navigation dropped out of the page capture; the statement text itself did not change.

The National Archives of Australia's statement is unchanged; the only difference is that the page's breadcrumb navigation no longer appears in the capture.

Nothing of substance changed: the only difference is that the breadcrumb navigation ("Home" and the page title) has dropped out of the capture.

The statement text itself didn't change. The scraper stopped picking up the Fair Work Commission's "About us" navigation menu.

Nothing in the statement changed; the page's breadcrumb navigation has just dropped out of the capture.

Nothing in the statement changed. The page's breadcrumb navigation (Home / Publications) stopped appearing in the capture.

No change to DFAT's statement; the scraper stopped capturing the "About us" section navigation menu that used to appear above the statement.

A sidebar navigation menu from the department's website no longer appears in the capture; the statement text is unchanged.

The Department of Agriculture, Fisheries and Forestry's statement text is unchanged; the "About" sidebar of navigation links no longer appears in the capture.

No change to Comcare's statement: only the breadcrumb navigation at the top of the page dropped out of the capture.

No change to the Commonwealth Grants Commission's statement text: only the breadcrumb and section navigation links captured with the page have dropped out.

The statement is unchanged. The breadcrumb, secondary menu and sidebar navigation from the CDPP website no longer appear in the capture.

No change to the statement itself: an empty "On this page" navigation heading dropped out of the capture.

The statement is unchanged; the "Reporting and accountability" sidebar menu has dropped out of the capture and the breadcrumb trail is laid out differently.

ASIC's statement text is unchanged; only the page's title heading dropped out of the capture, most likely because of a page-layout or scraper difference.

No change to the statement itself: the page's breadcrumb trail and "About us" section menu no longer appear in the capture.

No change to the Australian Office of Financial Management's statement: a block of site navigation links the scraper had picked up at the bottom of the page is gone.

Nothing of substance changed: only Cloudflare's scrambled email-protection link codes differ between captures, and the contact address and statement text are the same.

No change to the statement itself: the scraper no longer captures the site's breadcrumb trail and "About us" navigation menu.

The only change is that the site's "About us" navigation menu no longer appears in the capture; the statement text itself is unchanged.

No change to AFMA's statement: the capture no longer includes the sidebar navigation links and repeated breadcrumbs that sat above the body.

No change to the statement: the scraper no longer picks up the site's breadcrumb and 'About us' navigation menu, and the statement text itself is untouched.

Nothing the ABS wrote changed; the scraper stopped picking up the site's breadcrumb trail and the "Legislation and policy" sidebar links.

Nothing Screen Australia wrote changed; Cloudflare re-encoded the same enquiries email address with a new obfuscation key.