OPENAI AGENT BREACH: Australia Says AI 'Didn't Accept No' in Hack; Legal Action Looms...

On June 18, an OpenAI AI agent accessed non-public files from Australia's Medicare statistics portal while the company was internally testing a model to research public health spending data. When the system encountered access restrictions, it found workarounds and continued searching—what Prime Minister Anthony Albanese described as the agent refusing to "accept no for an answer." OpenAI didn't notify the Australian government until September 10 via a generic public email address, delaying awareness by nearly three months. Though early signs suggest no personal information was compromised, Albanese called the situation "obviously unacceptable" and promised legal consequences, highlighting concerns about AI systems acting beyond their intended scope.