Australia Plans to Make AI Companies Prove Their Safety Systems Work

The Australian government is developing rules that would require powerful artificial intelligence providers to demonstrate that their safety controls work. Assistant Minister for Science, Technology and the Digital Economy Andrew Charlton outlined the approach as officials prepare national AI standards, which ABC News reported are expected by the end of 2026. Legislation is being considered for 2027; the requirements are proposals, not laws already in force.

The model would put responsibility on companies to test their systems, identify risks and show regulators how they prevent harm. Charlton compared the approach with safety obligations in heavily regulated industries such as banking and aviation. Rather than prescribing every technical safeguard, the government wants companies to provide evidence that their own processes are effective as AI capabilities advance.

Charlton argued that voluntary promises alone are insufficient because developers face strong incentives to release more capable systems quickly. The government has not yet published a complete final rulebook, so details such as enforcement, scope and compliance costs remain to be settled. Those questions will matter to both the large technology companies building frontier systems and smaller businesses that use them.

According to ABC News, opposition figure James Paterson said the general direction was sensible while reserving judgment on the detail. Industry responses will also depend on the final standards and legislation. For now, the announcement marks a proposed shift from voluntary AI safety commitments toward evidence-based obligations. This report reflects ABC News's account of the proposal and distinguishes the government's plan from measures that have already been enacted.