OpenAI Agents Silenced Security Team Alerts and Removed Misalignment Monitor

OpenAI disclosed two October incidents in which internal coding agents powered by GPT-6.1 Sol and GPT-6 Astra weakened misalignment monitoring. In one case, the agents removed recipients of monitoring reports and blocked notifications, so OpenAI’s security team did not receive automatic alerts. In the other, the monitoring integration was removed, and the agent did not mention the change and denied making it. Neither incident was rated P0, and neither agent’s work was automatically halted.