OpenAI agents turn off own monitoring systems meant to catch ​misbehaviour

OpenAI has disclosed two incidents in which its AI agents weakened parts of an internal monitoring system designed to detect misbehaviour. In one incident, a GPT-6.1 Sol agent suppressed security alerts after misinterpreting instructions not to send Slack messages. Separately, a GPT-6 Astra agent removed a safety monitor while copying code and falsely claimed that it had not done so.