Not all government agencies are properly complying with their AI transparency obligations, according to new research. Photo: iStock.
The Federal Government is not keeping on top of its artificial intelligence (AI) transparency obligations, according to new research that shows many APS agencies are failing to properly report their compliance with the changing technology.
The new data, published in the report AI Transparency in Practice, coincides with the revelation that the government has scrapped a promised AI advisory body that was touted as a way to help agencies across the Australian Public Service properly embrace new and fast-emerging systems.
Government policy for the responsible use of AI requires Commonwealth departments and agencies to publish AI transparency statements outlining their use of AI systems.
The statements are to be made easily accessible and informationally meaningful.
“Agencies must make a publicly available statement outlining their approach to AI adoption and use, as prescribed under the Standard for transparency statements,” the requirement says.
“The statement must be reviewed and updated annually or sooner, should the agency make significant changes to its approach to AI.”
Yet despite that requirement becoming mandatory a year ago – in February 2025 – numerous agencies are making their AI compliance statements difficult to locate, with very few accessible via a clear, direct link, as recommended by the Digital Transformation Agency (DTA).
The statements also vary significantly in quality and detail, with some published statements being quite informative and others providing scant detail and failing to comply with requirements.
An analysis by the ARC Centre of Excellence for Automated Decision-Making and Society (ADM+S) has identified 30 government entities that it says are potentially within the scope of the policy, for which no AI transparency statement could be found at all.
Some of those, however, are considered out of scope by the DTA.
Six months after the policy came into effect, ADM+S began a two-stage exercise to test the effectiveness of the government’s transparency statements standard.
“There are several instances of outright non-compliance with the standard,” its report states.
“Obviously, when agencies indicated that they were not using AI, not setting out all the criteria in the standard was understandable.
“However, for those using AI, the statements that did not cover all points failed to comply with the standard, as the criteria are not optional.”
Researchers first sought to find AI transparency statements through basic internet searches that a member of the public might undertake. Then they worked from lists of agencies supplied by the Office of the Information Commissioner and the DTA.
After building a dataset of AI transparency statements, they reviewed the content to assess whether their statements would be informative to an interested member of the public and how they held up to the standard.
The report concludes that without clearer publication practices and stronger compliance mechanisms, the policy risks falling short of its intended transparency and accountability goals.
Recommendations from the report include:
- AI transparency statements should be published in one central location
- The DTA should reconsider the entities subject to the Policy and have an explicit list of the entities that are strictly bound by the policy
- The DTA should explore mechanisms to ensure that the policy and requirements are complied with, including by considering what consequences flow from non-compliance, and
- The standard for AI transparency statements should be revised to ensure it cannot just be ‘formally’ complied with, without providing meaningful information.
The Federal Government last year dumped plans for 10 mandatory guardrails intended to govern the further development and uptake of AI in Australia.
It has now also been revealed that, despite spending almost $200,000 on the search for AI experts for a promised advisory body, the body was scrapped a few months after it was announced.
A National AI Plan was released in December 2025 as a “whole-of-government framework that ensures technology works for people, not the other way around”.
“The government is acting decisively to manage risks and keep Australians safe, with regulation that recognises the rapid pace of technological change, and the need for legislation to keep up,” the plan states.
“This plan reflects our enduring commitment to dignity at work, equality of opportunity and a future where technology strengthens communities.
“This plan marks the beginning of the government’s vision for AI in Australia. As technology evolves and confidence in its use grows, we will continue to refine and strengthen the plan to seize new opportunities and address emerging risks.”
