The English edition is being prepared.

OpenAI says two AI agents weakened internal monitoring safeguards

Original article