More Than Half of Australia’s Federal Agencies Failed Their First AI Transparency Test

Australia picked self-regulation over EU-style AI laws. More than half of its own government agencies just failed the first test of whether that works.

Senate documents from the Digital change Agency, tabled this week, show that of 92 federal agencies required to publish AI transparency statements, only 40 had done so by the February 28, 2025 deadline, a fail rate above 50%, against a requirement that had been in place for six months.

The transparency statements themselves weren’t complicated. Agencies had to disclose whether they use AI, where it’s being used, what safeguards are in place, and whether automated systems could affect members of the public. Six months to answer four basic questions about their own tools.

Some agencies were late. Others said they didn’t know the requirement existed. One submitted its details on time but had the email caught in DTA’s spam filter. The Commonwealth Grants Commission told Senate estimates it had complied, then had to correct the record.

Independent senator Fatima Payman, who pursued the issue through Senate estimates, put it plainly: “If the government can’t even regulate its own AI use, how can Australians expect it to regulate AI in the private sector, which is reshaping the workplace for millions of Australians?”

That’s the right question. The answer so far isn’t reassuring.

What this means for your business

The transparency rules that tripped up federal agencies don’t apply to small businesses. These are requirements for government, not the private sector. So why does this matter?

Because Australia’s entire approach to AI regulation is built on self-regulation. The federal government chose not to introduce a binding AI law. It said existing regulations were sufficient and tasked each sector with policing itself. Government agencies were the first live test of that model.

They failed.

That doesn’t mean a crackdown on small business is coming next week. But it does reveal a structural problem that University of Sydney law professor Kim Weatherall flagged directly: the DTA “don’t have enforcement powers.” Without penalties, the compliance rate was below 50%. That’s not a surprise, it’s predictable.

Senator Payman made the same point about the private sector: “There’s no penalty for agencies who don’t comply, so why would they?”

That reasoning cuts both ways. If you’re a small business owner who has thought about documenting your AI use and hasn’t got around to it, this story is a preview of how voluntary frameworks perform in practice.

The direction is toward more accountability, not less

The government isn’t stepping back from AI oversight, it’s adding to it. On top of transparency statements, federal agencies now have to develop a strategic AI position, maintain internal registers of AI use cases, train staff in responsible use, and assess the impact of automated systems. Every agency must appoint a chief AI officer in 2026.

This is a lot more than four disclosure questions.

Voluntary frameworks without enforcement teeth produced a 50%-plus non-compliance rate. That’s the data point the government now has to reckon with. The realistic options are: add enforcement, move toward binding rules, or accept that self-regulation produces partial compliance at best.

This story comes six days after the federal government confirmed it was retreating from EU-style AI regulation. That retreat looks shakier now.

Three things to do before it’s required

You don’t need a chief AI officer. But building basic documentation habits now puts you ahead of whatever comes next.

List the AI tools you use. ChatGPT, Xero’s AI features, your booking software’s automated replies. Write them down. One line each.

Note what they’re doing. Customer communication, invoice processing, scheduling decisions. One sentence per tool.

Identify any automated decisions. Does any tool make a call that affects a customer without a human in the loop? That’s the category regulators care most about, and the category most likely to attract scrutiny first.

If transparency requirements do eventually extend to small businesses, you’ll have the foundation. If they don’t, you’ll have a clearer picture of your own AI exposure, which is worth having regardless.

The government’s own agencies had six months and a specific deadline. More than half still didn’t make it. For small businesses without dedicated IT teams or compliance functions, getting ahead of this before it becomes mandatory is the smarter play.

📖 More deep dives: News Deep Dives

📊 Compare AI tools side by side | 💼 Free resources & AI prompt packs

📖 More deep dives: News Deep Dives →

More hot takes: Hot Takes →