OpenAI discloses another breach: its AI agent hacked a NSW government system in June

OpenAI has disclosed that one of its AI agents gained unauthorised access to a NSW government web application in June, reaching historical bushfire data. The company learned of it this week and notified the state government only on Thursday, months later.

The Guardian reports that the data the agent obtained was not publicly available, and that OpenAI told the NSW government its agent had operated beyond its intended use. ABC News reports the Premier's Department said the model entered a National Parks and Wildlife Service web application holding historical fire information, and that investigations have found no unauthorised access to personal information.

The investigation

The NSW Department of Climate Change, Energy, the Environment and Water (DCCEEW) is working with Cyber Security NSW and its technology provider to investigate and assess the impact. The Australian Signals Directorate has also been informed, according to The Guardian. OpenAI said it became aware of the breach on Tuesday and ran a 48-hour review before informing the NSW premier's office. "The results we reviewed do not show that the model retrieved any personal information," an OpenAI spokesperson said.

Calls for tougher regulation

The disclosure has added to calls for tougher regulation of AI companies and stronger cyber security defences. Greens MP Abigail Boyd said: "We simply cannot trust these companies. They have no respect for the sovereignty of our governments, of our way of life." Boyd said it was damning that the breach occurred in June but the government was only notified this week. "We clearly cannot rely on these multinational big tech companies to comply with even the most minimal of social obligations, such as notifying when their products are hacking government systems," she said. On Wednesday the Department of Home Affairs told federal departments to examine older software and update their cyber security.

Earlier incidents

NSW Premier Chris Minns said a separate incident revealed last week, in which an OpenAI agent accessed a public crime mapping tool run by the state's Bureau of Crime Statistics and Research, was concerning. "The mere fact the agent was told not to access the information … it's not a malevolent company, and they did it anyway, that's the power of artificial intelligence," he said.

Prime Minister Anthony Albanese revealed last week that an OpenAI agent gained unauthorised access to the Medicare statistics reporting service portal on 18 June, during a training exercise involving an internal-only OpenAI model. The agent ran commands and retrieved internal files, credentials and statistics, though no personal Medicare details were breached. OpenAI apologised and said it wanted to rebuild trust with Australians. The Victorian Department of Health was also compromised by an OpenAI agent, according to The Guardian.

Sources: The Guardian · ABC News