For agencies
If you are drafting or updating your agency’s AI transparency statement, you can check what it has to cover, how many statements already cover each item, and examples from other agencies. Find your statement’s report card in the agencies list.
What a statement has to cover
Three documents set the content. The DTA’s Standard for AI transparency statements lists eight things every statement must contain. Version 2.0 of the AI policy, in force since 15 December 2025, adds obligations that statements are expected to report on. The APS AI Plan separately asks each agency for a Chief AI Officer. Who’s who explains how they fit together.
The counts are of the 121 published statements as of 13 Oct 2026, as read by a language model. For the policy and Plan items, the examples cover the item mostly in their own words.
| Item | Asked by | Statements that say so | Examples |
|---|---|---|---|
| Intentions behind AI use | Standard | 115 of 121 | |
| Use classified by DTA usage pattern or domain | Standard | 111 of 121 | |
| Public-facing use addressed | Standard | 100 of 121 | |
| Monitoring and protection measures | Standard | 115 of 121 | |
| Compliance with the policy | Standard | 110 of 121 | |
| Compliance with legislation | Standard | 98 of 121 | |
| Date last updated | Standard | 108 of 121 | |
| Public contact | Standard | 110 of 121 | |
| Review cadence | Policy v2.0, annually, and after any significant change | 89 of 121 | NCC AFP IGIS |
| Accountable official | Policy v2.0 | 105 of 121 | AFP ATSB HSRA |
| Strategic position on AI | Policy v2.0, due 15 June 2026 | 16 of 121 | NFSA HSRA ANAO |
| AI use-case register | Policy v2.0, due 15 December 2026 | 32 of 121 | AFA ANAO FINANCE |
| Staff training | Policy v2.0, due 15 December 2026 | 55 of 121 | NHMRC IGIS ATSB |
| Chief AI Officer | AI Plan, due July 2026 | 39 of 121 | IGIS IPA ANAO |
Statements that report every item as done (1): Aged Care Quality and Safety Commission
How the tracker reads your statement
The tracker records what the page says, not what your agency does. If your agency has a use-case register but the statement doesn’t mention it, the report card will say “not mentioned”. Plain, explicit wording makes errors less likely:
- name each role by its job title, and say which one is the Chief AI Officer and which the accountable official, even if one person holds both
- put a last-updated date on the page, and say how often the statement is reviewed
- say which version of the policy you comply with
- when you have done something an earlier version promised (appointed the officer, published the strategy), say so, so the promise reads as kept rather than quietly dropped
Where the statement lives matters too. The tracker follows one address per agency, so moving the page breaks the record until we find it again. A web page reads more reliably than a PDF, and a page that only links to a PDF is read as a signpost, not a statement. Pages behind bot protection (Cloudflare, Imperva or Akamai challenges) often can only be checked by hand, so they are checked less often.
Borrowed wording is fine
Each statement page shades the passages it shares with other agencies, and gives an originality score. The score is not a grade: the Standard asks for some of the same wording everywhere, and following it closely lowers the score. The shared wording page shows which passages agencies share. Describe your agency’s AI uses and limits in terms specific to its work.
If we have read your statement wrongly
Every reading is made by a model and can be wrong, and each sits beside the text it came from so it can be checked. If your statement’s report card, summary or change history misstates what it says, email ben.swift@anu.edu.au or open an issue on GitHub, naming the statement and the reading. We will check it against the text and correct it or explain it. The same address takes earlier versions we don’t have.