'The world needs to slow AI down' Anthropic co-founder tells BBC

The interviewer opens by reading Evan Hubinger’s line back to Anthropic: we do earnestly believe AI could kill all humans, exclamation mark. Clark, the company’s co-founder and policy lead, has to answer for a number he did not give, two days after his CEO argued the frontier should be paced and five days after a departing researcher said both Anthropic and OpenAI were gambling with everyone’s lives.

This is the company speaking in its own voice about the figure the rest of the week was spent arguing over, which makes it the closest thing to an official position on it. Worth watching next to the All-In Summit interview, where the same number is called made up.