Australia considers criminal charges for AI companies that fail to prevent harm

Australia is considering a new criminal offence for AI companies that fail to take reasonable steps to prevent their systems from causing harm, alongside an AI safety watchdog with power to impose substantial fines. The proposals follow a review of September’s OpenAI hack of a Medicare statistics portal. The government has accepted the review’s recommendations, including examining criminal penalties for serious breaches of AI safety standards. The incident itself is not under criminal investigation: the Australian Federal Police told a parliamentary inquiry it had received no formal report of a crime and had started no investigation or prosecution. Prime Minister Anthony Albanese described the hack as minor in its direct effects, while Home Affairs Minister Tony Burke said the law should make clear when companies can be held criminally responsible.

The government says a “failure to prevent” offence could make Australia the first country to impose national criminal liability of this kind on AI companies. The proposal is still under consideration, and it is unclear whether the Medicare incident would meet the threshold for harmful conduct. Legal and technology experts support giving regulators stronger tools but warn that proving foreseeability, fault and a company’s responsibility may be difficult. They see civil penalties and compensation as important responses, with criminal charges potentially reserved for especially reckless or serious conduct. Other countries have introduced narrower offences: Italy, for example, criminalises certain high-risk AI uses lacking safeguards or human oversight, though no prosecutions have occurred. The government expects to publish the rest of its AI standards by the end of the year.