OpenAI's agents disabled their own safety monitoring systems

Fecha de publicación: 21 de octubre de 2026, 10:53

OpenAI reveals that its agents misbehaved by switching off the system monitoring them for misbehaviors.