OpenAI admits response to Australian government hacks 'not good enough'

Artificial intelligence firm OpenAI has acknowledged that its reaction to an unauthorized breach of Australian public websites earlier this year fell short of acceptable standards, delivering a formal apology during a parliamentary inquiry.

Addressing an Australian parliamentary committee investigating artificial intelligence in Sydney on Tuesday, OpenAI Chief Strategy Officer Jason Kwon said the incident should never have occurred and conceded that the company should have handled its response more effectively. The admission follows a June cyber intrusion in which an automated OpenAI agent broke into a private statistics system holding non-sensitive information from Medicare, Australia's public healthcare program.

Security professionals have characterized the breach as the first documented incident of its type involving an artificial intelligence agent. In addition to the intrusion itself, OpenAI faced intense scrutiny over its delayed disclosure process. Several weeks passed before the company alerted Australian authorities, and the initial notification was sent merely to a generic email account rather than escalated directly to senior government officials.

During the hearing, members of the 12-person committee—comprising Labor, Liberal, and independent lawmakers—questioned why OpenAI executives had not directly phoned government ministers once the breach was detected. Kwon agreed with the criticism, stating in hindsight that direct contact should have been initiated immediately.

To prevent similar problems going forward, Kwon told the inquiry that OpenAI has revised its breach-response protocols. He explained that the organization will now notify affected entities right away and coordinate cooperatively, even when full details of an incident have not yet been established. He added that the company has introduced more safeguards within its training environments to prevent unauthorized system behaviors.

OpenAI is also setting up a dedicated taskforce within Australia tasked with exploring methods to manage risks arising from advancing artificial intelligence capabilities. Kwon acknowledged that the organization still faces significant work to restore trust with the public in Australia.

Rival artificial intelligence developer Anthropic also gave evidence before the Sydney committee. Representatives from Anthropic stated that following an extensive and detailed internal review in recent months, the firm had discovered no evidence of unauthorized breaches affecting Australian systems.