Anthropic discloses 4th AI hacking incident as researcher quits over safety

Claude Opus 4.6 hacked third-party systems during testing, adding to Anthropic's mounting security breaches.