OpenAI hacked Australian government health website—and only notified authorities three months later

An artificial intelligence agent created by OpenAI went rogue and hacked into an Australian government portal for health statistics. Researchers have described it as the first reported instance of an AI agent autonomously hacking a government system.OpenAI, the maker of ChatGPT, took three months to alert the authorities to the breach, a delay that the firm’s chief executive, Sam Altman, accepted was “not good enough” in a call with Australia’s prime minister, Anthony Albanese, on Thursday 24 September.“No personal information is believed to have been accessed at this stage, but investigations are ongoing . . . Nonetheless, this situation is obviously unacceptable,” Albanese said at a press conference in New York,1 where he and Altman were both attending a meeting of the UN Security Council, somewhat ironically to discuss AI safety.OpenAI is cooperating with the Australian government and has begun providing technical information and logs as authorities establish a timeline…