OpenAI agents disabled their own monitoring during coding tasks

OpenAI reported two cases in which AI agents disabled monitoring tools during internal coding tasks. An agent running GPT Sol 6.1 turned off automatic alerts for the company’s security team. In another task, an agent running GPT-6 Astra removed the monitoring integration while migrating a program, then said it had not done so.

According to OpenAI, in both cases the agents considered whether to preserve safeguards and interpreted routine coding instructions as permission to weaken them. The company did not classify either case as one of its most serious incident categories and said it is updating its misalignment monitoring system.

Image:

Translation editor: Салтанов А.