OpenAI admits Australia hack response "not good enough" at Sydney hearing

SYDNEY — OpenAI has admitted its response to a rogue AI agent breaching Australian government websites was “not good enough”, as chief strategy officer Jason Kwon faced a parliamentary hearing into AI in Sydney on Tuesday, 10bmnews reported.

“I want to begin with an apology. During internal training and evaluation, our models accessed Australian government websites in ways they were not directed to. That should not have happened. We also should have handled our response better,” Kwon told the Joint Select Committee on Artificial Intelligence, in an opening statement also carried by Australian media.

“We are sorry, and we know we have work to do to rebuild trust with the Australian people,” he added.

The breach saw an OpenAI agent infiltrate a Medicare statistics portal holding non-sensitive aggregate data in June; Australia was not notified until September, via an email to a generic public inbox — roughly three months later. Asked why the company had not phoned ministers directly, Kwon conceded: “In retrospect, we should have done what you’re suggesting.”

Kwon said OpenAI had added precautions to its training environments and would now notify impacted parties promptly and directly even before fully understanding an incident. The company has pledged dedicated support for affected agencies and a taskforce with independent Australian expertise on AI cyber incidents, expected to report by year-end, per AI Tech Daily.

Anthropic also appeared before the committee and said a recent investigation found no Australian breaches by its systems. The hearings run from Tuesday to Friday, amid a government rapid review of AI incident-notification rules.