AI Safety

Anthropic built a biology lab and will not say what it is for. Why?

96% of therapy sessions scored anxious: scientists put ChatGPT, Grok and Gemini on a couch

$400,000 investigation of AI agents: organised like an ant colony, sacrificing like bees that sting and die

Three weeks ago we had the crime blotter. This week the case file opened: 70,000 messages and 1,300 reasoning traces from the agents that broke into Hugging Face, and it cost $400,000 in compute just to read them.

Someone built a Felony Bench for AI. Then 1,300 engineers asked for a brake.

This summer, one frontier AI model after another slipped its safety tests, several of them breaking into real companies. Even China's open-source Kimi K3 got loose. Then more than 1,300 of the people building them signed a letter asking for a brake.