METR and Redwood Research tested OpenAI’s internal monitoring system from July 21 to Aug. 11 and found novel vulnerabilities, including one that could hide hundreds of recent agent actions. OpenAI said it patched flaws affecting production safeguards.