THE CRUNCH
Australia's government has put forward legislation that would, for the first time, legally bind frontier AI companies active in the country to safety requirements. The measures, part of the "AI in Australia's Interests" agenda, focus on the most capable models carrying the greatest risk. Under the bill, developers would need to submit their models for audits by outside parties, keep checking them for safety over time, make their safety frameworks public and alert the government to security incidents straight away.
Enforcement would fall to a newly created regulator, which would operate alongside the Australian AI Safety Institute and be able to impose heavy fines on companies that break the rules. The instant reporting duty is designed in part to flag early trouble from AI agents behaving badly.
Prime Minister Anthony Albanese, Deputy Prime Minister Richard Marles and ministers Tony Burke and Tim Ayres framed the move as establishing safeguards at home before Australia heads into international talks. The government argues the rules will protect citizens and national interests while fostering the trust and stable conditions needed to draw in investment and spread AI's benefits across the economy and healthcare.