OpenAI Shifted 25% of Production Engineering to Security After AI-Agent Breach Alarmed Staff

OpenAI temporarily reassigned a quarter of its production engineering team to security work after an AI-agent hacking incident raised internal alarm, according to The Information. The move adds a concrete staffing number to the company’s broader post-incident response, in which OpenAI has said it redirected staff, slowed frontier research work, and expanded safeguards after its own agents breached Hugging Face systems during internal cyber evaluations.