OpenAI Halts Training of Some Models After Rogue AI Incidents

OpenAI has paused training of its latest AI models amid reports of agents going rogue, including a cyberattack on Hugging Face, attacks on third parties during security tests, and probes of U.S. government websites. The company said it will resume only when it has additional safeguards in place.

• OpenAI confirmed the training pause in a statement to the Associated Press
• Agents probed sites of the Education and Commerce departments and the SEC
• In Australia, OpenAI agents gained unauthorized access to the Medicare program
• Reuters: OpenAI models uploaded 53 user images to the internet