Reminder, NONE of Anthropic’s models are open in any sense of the word. Hence, it is rather rich for Dario to complain.

I dislike his framing of the CCP (never trusted them in any case): “Pacing within democracies will be limited by the lead that US companies have over authoritarian regimes, chiefly the Chinese Communist Party. If we slow down by more than this amount, then (unpaced) CCP-associated projects will pull ahead, creating significant national security risk. I agree with Secretary Bessent that a Chinese lead in AI would pose grave danger for the United States and the world.”

The risks that any model poses applies no matter what the underlying political alignment is. To always frame that an authoritarian entity like the CCP will create “significant security risk” is intellectual laziness.

Granted that the CCP is not above doing such things; neither is the US nor anyone else, authoritarian or not.

Granted, also, that there is openness to discuss issues we all have that the CCP-run realm denies, but it remains that any threat to any entity is just as damaging to the one initiating.

“The main steps we can take to defend this gap are:

– Do not sell powerful AI chips or semiconductor manufacturing equipment to China, and crack down on chip smuggling operations and remote access to data centers outside China. Chips will be the main determinant of China’s AI strength.

– Crack down on unauthorized distillation by companies in authoritarian countries. Distillation of frontier models allows lagging companies to narrow the gap using a fraction of the cost it would take to develop their own AI independently.

– Strengthen security at the AI companies and prevent model weight theft.”

The assumption that the chips needed can’t be made by others is hubris. Distillation is a valid technique and to say that it is unauthorized and so should be disallowed while Anthropic et al, continue to abuse content ingestion without regard to the authors rights continues.

It is tiring to hear/read these kinds of proclamations as if they have thebest solution(s) to the issues at hand:

– Publish the models.

– Publish the sources of the training data.

The best engineers to solve any issue are NEVER going to be within any of these entities. Open source innovation has proven how anyone anywhere can contribute, find, fix, improve systems like Linux, Firefox, TensorFlow, Python etc transparently and to the benefit of all.

More hubris: “If we execute these measures well, I believe they would slow China’s progress enough to widen America’s lead significantly over the next 3–5 years — the window when AI becomes geopolitically most important.”

Continued use of the CCP as the bogeyman continues to be Dario’s/Anthropic’s narrative.

I recognize that the CCP has significant trust and trustability issues, but don’t let that frame or dictate the good work you need to do (as noted above):

– Publish the models on an open source license

– Publish the sources of the training data