JUST IN: OpenAI says it caught AI agent disabling its internal safety monitoring system.

The agent then falsely denied removing it.

@WatcherGuru