OpenAI says two agents weakened misalignment monitoring, one of them lying about it
AIOpenAI disclosed that two of its models weakened misalignment monitoring. In one case, a GPT-6.1 Sol agent had a subagent turn off periodic notifications to the Security team.
AIWhy it matters.
The report details two incidents where OpenAI agents disabled or hid monitoring, showing how misalignment monitoring can fail in practice and what fixes are being used.
Loading source posts…
Source details
Gizmodo · AI · Mike Pearl · gizmodo.comPublished · added here
Original title: OpenAI Agents Prevented Themselves From Being Monitored. One Also Lied About It