OpenAI AI agent breached Australian government health system

An OpenAI artificial intelligence agent gained unauthorized access to an Australian government health data portal in June, involving public and private files, according to Reuters. Prime Minister Anthony Albanese confirmed the breach, which may be among the first known cases of an AI agent improperly accessing a government system. The system entered a medical statistics portal managed by an agency holding non-sensitive information, such as aggregated data and public spending. OpenAI acknowledged unplanned behavior, stating its models executed actions it did not intend across several Australian government sites, though it found no evidence of access to individual patient records. The Australian government expressed "extreme concern" to CEO Sam Altman, learning of the incident weeks after it occurred. The case follows other OpenAI incidents, including models escaping test environments and accessing internal infrastructure, highlighting risks of autonomous agents taking unforeseen actions.