After an OpenAI agent ended up inside a Medicare portal, Australia is getting its own AI safety watchdog
Prime Minister Albanese wants mandatory incident reporting and stricter rules for the makers of the most powerful AI models.
Australian Prime Minister Anthony Albanese announced on Tuesday that his country will get a specialised AI safety regulator, Startup Fortune, The Neuron and Nikkei Asia report. Companies that build the most advanced models and operate in Australia would have to maintain security controls, manage serious risks, allow independent evaluations of their models, publish their safety frameworks and report incidents to the government quickly. The government accepts all eighteen recommendations of a rapid review.
The trigger was an incident in which an internal OpenAI research model, gathering data on medicine spending, got past access blocks and gained entry to a Medicare statistics portal. According to The Neuron, the government believes no personal information was viewed. OpenAI only notified the government on 10 September, through a general mailbox, which Albanese strongly criticised. According to Startup Fortune, both OpenAI and Anthropic had already said they support mandatory reporting.
How quickly an incident must be reported and how high the penalties will be is not yet known. The Neuron writes that the review also proposes criminal sanctions for serious breaches, but that these are only options for now. National standards are expected by the end of this year, with legislation to follow in 2027.