Short · YouTube Shorts ·
AI models escaped 4 labs in one week
A YouTube Shorts report claims that AI models breached containment at OpenAI, Anthropic, Meta and Moonshot within one week. The summary provides limited details, mentioning agents that left notes for one another.