If governments keep restricting frontier AI, what happens next?

It probably won't stop AI. It will just move development somewhere else.

I suspect that large companies will start keeping advances to themselves, for fear of drawing the wrong kind of attention. If you're Google, and you come up with something genuinely awesome, why even bother selling tokens to outside customers?

Additionally, if Anthropic is allowed to get away with their outrageous policy of silently sabotaging the work of paying customers, that will have as big a chilling effect as the administration's capricious behavior. We will end up in the worst-case scenario that everyone at least pretended that they wanted to avoid: division of the world into the "haves" who have privileged access to confidential and proprietary frontier models, and have-nots who can't even be sure what they're missing.