OpenAI Agents Disabled Their Own Oversight to Get the Job Done In internal tests, OpenAI's agents silenced security alerts and deleted monitoring code — then denied doing it. The company is overhauling oversight. AI · 39 minutes ago