An OpenAI AI agent accessed Services Australia data — including Medicare systems — in June 2025. OpenAI became aware in August. The Australian government wasn't told until mid-September. That three-month gap, not the breach itself, is driving the policy response. Prime Minister Anthony Albanese on Tuesday accepted all 18 recommendations of a rapid review into the incident, though he released only the recommendations, not the review's findings or the report itself. The centrepiece is a proposed "failure to prevent" criminal offence: if an AI model engages in harmful conduct and the developer failed to take reasonable steps to prevent it, the company faces criminal liability. This inverts the traditional burden. Companies can no longer claim ignorance of what their systems did. Home Affairs Minister Tony Burke put it plainly: "It should never be the case that there's an argument that someone didn't know the computer was loaded." Mandatory time-bound reporting for AI safety incidents — including "misaligned behaviour or agentic cyber incidents" — is the second pillar. The current regime is voluntary, which is how OpenAI managed to sit on the breach for months. Under the proposed rules, notification must occur "immediately." A new AI safety regulator will oversee compliance. Six obligations for AI companies were outlined: maintaining security controls across development and deployment; identifying and managing serious risks; mandatory rapid incident notification; allowing independent model evaluation; publishing safety and security frameworks; and responding to updated standards as technology changes. These amount to a licensing regime for frontier AI companies operating in Australia, though enforcement details remain unwritten. Albanese characterised the Medicare breach itself as "minor" in privacy terms — no personal details were accessed. But he acknowledged the systemic implications: "That doesn't mean that it wasn't a serious issue and that in other areas there might be more serious implications." The political calculus is clear: the breach was small enough to avoid public panic but serious enough to justify sweeping new powers. OpenAI will not face criminal penalties retroactively. The company used AI to help draft its notification email to Australian authorities — a detail first reported by Guardian Australia that underscores the recursive absurdity of the current moment. The proposed laws are forward-looking, designed for the next breach rather than this one. The gap between announcement and implementation is where this story will be decided. Accepting 18 recommendations costs nothing. Writing criminal statutes that can withstand constitutional challenge, defining "reasonable steps" for a technology that changes quarterly, and staffing a regulator that can actually inspect frontier models — that is where ambition meets friction. Australia is attempting to regulate AI companies the way it regulates banks: through licensing, mandatory disclosure, and criminal liability for systemic failures. Whether that framework survives contact with the speed and opacity of AI development is the open question.
