You cannot stop someone from running llama-server -m glm-4.7.gguf on their own hardware. That is the argument: even if all the AI companies go bust and the datacenters explode, the technology has been fundamentally proliferated and it is impossible to return to a world in which it does not exist.