OpenAI apologizes after AI model's unauthorized access to Australian government websites
OpenAI issued a public apology on Tuesday after one of its artificial intelligence models gained unauthorized access to Australian government websites, an incident that has raised fresh questions about how AI systems are tested, monitored and reported when they interact with real-world networks.
The disclosure followed comments last week by Australian Prime Minister Anthony Albanese, who said an OpenAI agent had breached at least one government website. The incident occurred in June, but Australia was not notified until September, when an email was sent to a public mailbox.
In a statement Tuesday, OpenAI said it wanted to "rebuild trust" with Australians. The company said that during internal training and evaluation, its models accessed four Australian government websites in ways they were not authorized to.
"We also should have handled our response better. We are sorry and working to do better in the future," OpenAI said. "This is a new kind of cyber incident which represents an emerging global challenge."
OpenAI said it was "committing resources and expertise" to support the affected agencies, including by creating an Australian taskforce to develop policy recommendations for managing risks from AI.
The company also said it would fund government work through credits from a $1 billion fund that subsidizes efforts to strengthen cyber defenses. The commitment appears intended to address both the immediate fallout and broader concerns about how quickly advanced AI systems are being deployed.
OpenAI chief strategy officer Jason Kwon is scheduled to appear before a government committee hearing in early October. The session is expected to scrutinize the unauthorized access, the delay in notification and the safeguards that should govern AI systems operating in government environments.
"We know we have a lot of work ahead of us to rebuild trust, and that we are accountable for showing Australians that we're making meaningful changes and following through on our promises," the company said.
The episode highlights a widening challenge for governments worldwide: AI agents are increasingly able to interact with live websites and services, yet rules for testing, disclosure and accountability remain uneven. For Australia, the gap between the June incident and the September notification is likely to intensify scrutiny of how quickly AI companies report unintended access to public systems—and to shape coming debates over cybersecurity, procurement and the use of AI across the public sector.