Dozens of government agencies have failed the first test of policing their own use of AI after Australia withdrew from stricter European-style AI rules in favor of relying on existing regulators and agencies.
Last year, the federal government chose a more flexible model for regulating artificial intelligence, rather than the European-style approach of a single AI law with binding rules for high-risk AI.
The federal government said existing laws were sufficient, but it mandated that agencies police their own use of AI and then regulate the use of new technologies by their own industries.
But a document submitted to the Senate by the Digital Transformation Agency (DTA), the government-wide technology agency, showed the challenges with these approaches.
More than half of the government agencies checked by the DTA failed to meet initial transparency deadlines for disclosing how they use artificial intelligence, according to documents submitted to Congress.
Fatima Payman said the record proves that government agencies are not fulfilling even their basic duties. ( ABC News: Nick Haggerty)
Independent Sen. Fatima Payman, who pursued the issue through the Senate estimates, said the record proves the agency is failing even in its basic duties.
“If the government can’t even regulate its own use of AI, how can Australians expect it to regulate the private sector AI that is reshaping the workplaces of millions of Australians?” Senator Peyman told the ABC.
DTA Deputy CEO Lucy Poole acknowledged the challenges of regulating rapidly changing technology.
“The speed of change that AI is bringing to Australia’s public services means we need to be flexible and thoughtful to ensure it is applied meaningfully and responsibly where appropriate,” Mr Poole said.
Most government agencies missed AI publication deadlines
Ahead of a change in government direction following the 2025 federal election, DTA has set a policy requiring government agencies to share basic information about their use of AI.
DTA’s use of AI in government policy was aimed at building trust through transparency.
These transparency statements were intended to inform the public whether agencies are using AI, where it is being used, what safeguards are in place, and whether people could be directly affected by automated systems.
However, DTA records show that by 2pm on February 28, 2025, a deadline set six months after the policy took effect, only 40 of the 92 delegated institutions had reported to the DTA.
After publication of this article, DTA provided additional documentation showing that subsequent compliance inspections of agency websites found only 13 to be non-compliant.
Additionally, the DTA later reported to the Senate that 30 of the 92 delegated agencies had not met another new requirement for the agency to appoint someone responsible for the use of AI by the previous deadline of November 2024.
Late, unnoticed, email sent to spam
Some agencies were delayed. Others said they were unaware of this requirement or later revised their answers to Senate questions.
In one case, the Federal Grants Commission initially presented evidence that both deadlines had been met, but later acknowledged that this had not been the case.
Another agency, the National Competition Council, said in a Senate estimates hearing that it did not issue an AI transparency statement by February 2025 because it was not aware of the deadline.
It then said it was compliant “by extension” because it relied on the ACCC for information technology services and the ACCC issued its own statement in time.
The Organ and Tissue Authority (OTA) was initially treated as non-compliant by the DTA, but subsequent correspondence revealed that the agency emailed responsible official details on December 5, 2024.
My OTA email was getting caught in DTA’s spam filter. DTA later corrected the record and said OTA had complied.
Kim Weatherall, a law professor at the University of Sydney, said the first attempt at a transparency statement showed the challenges of assigning responsibility to institutions.
Professor Weatherall said: “This just adds to their challenge.”
She said the initial transparency disclosure was a “mixed bag”.
Some provided very detailed information, she said, while others provided little information about the type of AI they were using or what rules there were for its use.
But she cautioned against assuming that the initial efforts at transparency statements meant the agency wasn’t doing a good job managing the use of AI within the department.
“There is a good chance they are doing a good job internally while dealing with a rapidly changing situation.”
Professor Weatherall said:
Government raises expectations for AI
The DTA says all 94 government agencies are currently meeting their obligations regarding AI transparency and the appointment of responsible officials.
“As this was a new policy, the DTA worked closely with the empowered agencies to ensure they understood the requirements and delivered the expected outcomes,” Mr Poole said.
We also moved to make AI transparency statements easier to find.
Previously, Australians had to search individual government agency websites to compare statements about government use of AI. The DTA has now created a central registry of AI transparency statements.
“We will iterate on the registration to improve the user experience and the details included,” Poole said.
DTA Deputy CEO Lucy Poole acknowledged the challenges of regulating rapidly changing technology. (Provided by: DTA)
Government agencies are also being given new responsibilities regarding the use of AI in government.
In addition to transparency statements, agencies should establish a strategic position on AI adoption, maintain internal records of AI use cases, train staff on responsible AI use, and evaluate the potential impact of AI use.
The government’s AI for Public Services plan also expects each agency to appoint a chief AI officer in 2026.
This is a new role, distinct from the government agency’s compliance officer, and the AI officer is tasked with driving the adoption of AI within the agency.
Professor Weatherall said there were still structural weaknesses in the framework as the DTA was not an enforcement agency.
“They’re trying to do the right thing, like creating standards and guidelines, but they’re not regulators,” she said.
“They have no enforcement powers.”
Sen. Peyman said agency compliance issues have raised questions about the government’s ability to regulate AI outside of public services.
“Many government agencies have not met basic AI transparency requirements, some months after they were mandated,” she said.
“There are no penalties for agencies that don’t comply, so why penalize them?”
Finance Minister Katie Gallagher’s office did not respond to a request for comment.
