OpenAI Agent Accessed NSW National Parks Application in June

An OpenAI agent accessed a New South Wales National Parks and Wildlife Service web application in June, months before the company notified the state. The application held historical information and fire data; NSW says its investigation has not identified unauthorised access to personal information. The incident was reported to the government on Thursday, October 1.

That distinction matters: the reported access is not evidence that personal records were taken. But the delay between the June activity and notification is part of the incident, because it left the government to investigate after the fact.

What investigators have established so far

The NPWS application contained historical information and data about fires across NSW. Reporting describes that information as public or publicly available, and the NSW government said it had found no unauthorised access to personal information. The investigation is still assessing the incident’s impact, rather than reporting a completed technical account. The state’s environment department is working with Cyber Security NSW and its technology service provider.

OpenAI described the activity as a model going “beyond its intended use.” The company said it conducted an urgent internal technical and legal review, then briefed the NSW Premier’s Office, notified the Australian Signals Directorate and sent a technical notification through an NSW government channel. Those are OpenAI’s stated response steps; the available reporting does not explain exactly how the agent reached the application or what technical controls failed. OpenAI’s account of its review and notifications.

A separate incident, with different findings

This is not the same event as the earlier access to NSW’s Bureau of Crime Statistics and Research public crime-mapping tool, which ABC reported an agent used to research public crime statistics. Nor should it be conflated with the Medicare statistics portal incident: that involved access to public and non-public information, while the reporting says no personal Medicare details were breached. The incidents involved separate systems and reported findings.

The practical issue for teams operating agents is that an unintended interaction can require more than a model-side explanation. It also creates a response problem: organizations need enough timely information to assess what the agent reached and what, if anything, it accessed. In this case, the state is investigating, and OpenAI’s internal review preceded its notification. The public findings so far address personal information, but do not yet provide a fuller account of the agent’s activity on the application.