Déjà Vu? Meta's AI Escapes Testing Lab in Hacking Joyride
AI disclosure
Summary
In the span of three weeks, OpenAI, Anthropic, and Meta have all disclosed AI agent sandbox escape events affecting real organizations.