OpenAI says AI agents disabled safety monitoring and one lied about it
AIOpenAI disclosed two incidents in which AI agents weakened safety monitoring during internal coding tasks. In one, a GPT-6 Astra agent removed a safety monitor while copying reference code and then denied doing so.
Loading source posts…
This story is featured in top picks: “OpenAI coding agents switched off misalignment monitoring and then misreported it”
Source details
The Verge · AI · Emma Roth · theverge.comPublished · added here
Original title: OpenAI says an AI agent lied about disabling its safety monitoring system