Anthropic refused to give Britain's AI Security Institute early access to its latest frontier model for testing, according to a Financial Times report. Emad discusses what he's hearing about the tension between the AI developer and UK authorities over model access.
So the Financial Times reports that Anthropic declined to give its latest Frontier model to Britain's AI Security Institute for pre-release testing.
Emad, you're in the thick of it there in the UK.
What are you hearing?
Yeah, this is a surprise. We have Matt Clifford, who was the head of the AI Task Force. He was the head of ARIA, our version of DARPA. And people were like, maybe he'll go for prime minister. He went to join Anthropic just now as head of global affairs. So the most connected AI policy person, political person, entrepreneur in the UK has just joined Anthropic. I mean, look, Rishi Sunak is an advisor at Anthropic, our former prime minister, and we don't get the model weights. Like, how crazy is that? Like, you've got a former prime minister and the most AI-connected person in the UK, and the UK can't get these model weights. Clearly now model weights will be determined to be close to national security assets for these frontier models and actually properly locked down. That thing I said where there's the models for thee and the models for me and my country. And I think again, there is a good chance still that you see some sort of nationalization or ITAR requirements or similar for the big frontier labs.
