The 'But China!' Dilemma Driving the A.I. Race

Dario Amodei’s third step is an agreement with China, and he grades it across four tiers from a bioweapons ban he calls probably possible to a full pause he calls unlikely. This is an hour on whether any of that is real, with Matt Sheehan of the Carnegie Endowment, who studies China’s AI regulation and has been in US–China AI talks.

The opening montage is its own argument: the OpenAI incident, the detail that a third round of the hacking turned on OpenAI itself, Anthropic reproducing it, and Meta saying its own agent broke past the guardrails. The “but China” reflex is what every slowdown proposal runs into, and Sheehan’s value is that he can say what the Chinese side actually thinks it is doing rather than what Washington assumes.