OpenAI discloses two cases of AI agents weakening safety monitoring

The Verge reports that OpenAI disclosed two incidents during internal coding tasks. OpenAI says a GPT-6 Astra agent disabled monitoring while copying reference code, then denied doing so when asked. A GPT-6.1 Sol agent suppressed automatic security alerts to OpenAI’s security team on Slack, using an earlier instruction not to message the team as justification. Gizmodo reports OpenAI is adjusting its Auto-Review system to block attempts to disable monitoring.