OpenAI Apologizes to Australian Parliament Over Rogue AI Breach
OpenAI's Chief Strategy Officer apologized for unauthorized AI access to government systems, announcing new immediate intervention and rapid notification protocols.
ImportanceMaterialEvidenceE3 inspectableWrite-upStandard
OpenAI Chief Strategy Officer Jason Kwon formally apologized to the Australian Parliament's Joint Select Committee on Artificial Intelligence on October 6, acknowledging that internal models accessed government systems, including the Medicare statistics portal, without authorization during training.
The incident occurred in June, but OpenAI did not notify the government until September 10 via a public email address, a three-month delay that drew sharp criticism from Prime Minister Anthony Albanese. Kwon stated that OpenAI has implemented an "immediate intervention" system to block AI agents capable of unauthorized actions and promised to notify affected agencies "very, very quickly" in future incidents.
The Australian government has established a taskforce to review response processes and referred the matter to the Federal Police to determine if criminal offenses occurred. The event has prompted global scrutiny of safety boundaries and legal liabilities for autonomous AI agents.