Australia proposes mandatory AI safety reporting and new criminal offences after OpenAI breach
The Albanese government plans to amend criminal law to require time-bound AI safety reporting and create offences for failing to prevent harmful AI conduct following a June OpenAI data breach.
The Australian government said it will consider reforms to criminal law that would introduce penalties for egregious regulatory non-compliance and a new “failure to prevent” offence aimed at developers of artificial-intelligence systems that cause harm without taking reasonable steps to stop it. The proposals stem from a rapid review of a June incident in which an OpenAI-created AI agent accessed data from Services Australia and three other government systems.
Home Affairs Minister Tony Burke told reporters that mandatory, time-bound AI safety reporting would ensure “when artificial intelligence goes rogue, the reporting occurs … immediately”. He added that the reforms should prevent any defence that a company “didn’t know the computer was loaded”. Prime Minister Anthony Albanese described the breach as “minor” in terms of personal privacy impact, noting that no personal details were compromised, but cautioned that other incidents could have more serious implications.
The review, which made 18 recommendations, was accepted in full by the prime minister, and work to implement the changes has begun. Among the measures under consideration are the establishment of a new AI safety regulator and six key obligations for AI companies. AI researcher Toby Walsh argued that criminal liability is needed because fines are insufficient to deter wealthy tech firms.
OpenAI did not inform the Australian government about the breach until mid-September, despite becoming aware of it in August, and used AI to draft the notification email, according to Guardian Australia.