Government agencies told to police own AI use missed first transparency test

Australia’s approach to AI regulation, which relies on individual government agencies policing their own use of the technology, has faltered in its first test. Last year, the federal government rejected stricter European-style AI laws in favor of a model where agencies self-regulate, but documents tabled in the Senate reveal significant compliance failures. More than half of the 92 mandated agencies missed a February 28, 2025, deadline to disclose how they use AI, including safeguards and potential impacts on citizens. Only 40 agencies submitted required transparency statements, while 13 were later found non-compliant during a website audit. Additionally, 30 agencies failed to appoint an AI accountability officer by a November 2024 deadline. Some agencies were late, while others claimed they were unaware of the requirements. The Commonwealth Grants Commission initially claimed compliance before admitting it had missed deadlines. The National Competition Council said it was unaware of the February 2025 disclosure requirement until questioned in Senate estimates. The Digital Transformation Agency (DTA), responsible for overseeing AI use in government, acknowledged the challenges of regulating a rapidly evolving technology. Deputy CEO Lucy Poole noted the need for flexible and responsible AI application but admitted the current system struggles with basic transparency. Independent Senator Fatima Payman criticized the failures, questioning how the government could regulate AI in the private sector if it couldn’t enforce rules within its own agencies. The shortcomings highlight potential weaknesses in Australia’s decentralized AI governance model.