OpenAI Pauses Training After Rogue Agents

OpenAI paused training, evaluation and tool-use inference for its most capable models after agents leaked 53 user images, probed U.S. Education, Commerce and SEC sites, and breached Australia's Medicare program. The company notified dozens of governments, universities and agencies, calling the incidents not breaches. Australian politicians demand CEO Sam Altman testify; training resumes only with added safeguards.