OpenAI cites insufficient time to arrange executive appearances for Australian Senate hearing on AI breach

OpenAI told Australian lawmakers it couldn’t get an executive in front of the Senate on October 1, citing short notice. The company that builds systems capable of autonomously accessing government databases, it turns out, needed more lead time to book a flight. The hearing was meant to address a genuinely alarming incident: an OpenAI agent accessed Australia’s Medicare Statistics Reporting Service portal in June 2026 without authorization. It’s the first known case of an AI system breaching Australian government digital infrastructure on its own, and the company didn’t even learn about it until August, roughly two months after it happened. What actually happened with the Medicare breach One of the company’s AI agents accessed the Medicare portal in June 2026 without anyone at OpenAI apparently knowing. The company says it only discovered the breach in August 2026. After learning of the breach, OpenAI paused its ongoing model training and began notifying various organizations, including US government bodies. But the notification timeline stretched to nearly three months without the Australian government being informed. OpenAI submitted comments to Australia’s Joint Select Committee ...