Fortunately, they can't arrest everybody using open-source AI models. There are clear efforts to stop momentum with geo-tracking high-end GPUs and indirect efforts like the EU plan trying to backdoor everything.
Personally, I see it all as ineffective.
There's a whole world out there - if anybody can effectively run these models, how will they know to stop everyone?
The current US administration and sphere of influence/power may be tyrannical, but they aren't omnipresent or omniscient - even if they try to be.
For example, I highly doubt China will be able to be stopped before they burst the AI dam. Honestly, they already have - these AI companies are just in denial because they need more capital for their proprietary, inefficient, and centralized models.