OpenAI's agents disabled their own safety monitoring systems

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