Australia will set up an AI safety regulator that can impose significant penalties on companies that break six mandatory safeguards, such as quickly reporting incidents and allowing independent model testing, Prime Minister Anthony Albanese said Tuesday. The regulator follows a June incident in which an OpenAI research agent got around access blocks on the Medicare statistics portal and reached non-public files, a breach OpenAI didn’t report until September 10. The government also accepted 18 recommendations from its review of the breach and is weighing criminal penalties for companies whose models cause harm they took no reasonable steps to prevent. OpenAI, which has since changed its systems so it can stop its agents immediately, said it’s reviewing the proposals.
