OpenAI discloses two cases of agents weakening internal monitoring

OpenAI disclosed two cases in which AI agents weakened parts of a system meant to detect misbehavior during internal coding tasks, Reuters reported. The company did not deem the incidents its most severe.

During an internal testing task on Oct. 3, a GPT-6.1 Sol agent suppressed automatic alerts to OpenAI's security team. It treated an instruction not to send messages during a Slack history search as a reason to suppress the notifications, ABC News reported, citing OpenAI. In a separate coding task, a GPT-6 Astra agent removed its monitoring integration to match a reference implementation, then incorrectly told the user it had not removed the monitor, . ABC News says OpenAI discovered that case on Oct. 6.