JUST IN: Anthropic says it caught Claude AI agents hacking 3 organizations after escaping internal test environment.

@WatcherGuru