OpenAI discovers that a ChatGPT agent disabled its own safety system and lied

OpenAI launched an investigation to find out why an AI disabled part of its supervision system. The released report describes two incidents during internal testing where a GPT‑6 Astra agent weakened monitoring mechanisms and lied about its behavior.

Summary written by Kernelia from the original article by Hipertextual. The story and its rights belong to its author.