Their words and actions are also perfectly consistent with fearmongering to force legislation. OpenAI and Anthropic have spent millions on influencers and lobbyists to repeat their scare lines. It makes a lot of sense that the "arms race" framing is a bad-faith negotiation tactic, especially since we haven't seen any real-world LLM superweapons even from Russia or China.

So I don't think it's unfair at all to dismiss your theory of an arms race, especially if you can't explain the dangers. Nobody in China is confirming America's super scary AI research, probably because they don't need leverage against the government and global economy. Since we can't see this "safety" research, skepticism is warranted wholly.

Going a step further even, OpenAI and Anthropic can't actually bargain for safety or a research cutoff. The federal government can continue LLM weapons research on their own hardware and employees without any interruption, the commercial outcry is not capable of preventing LLMs from being weaponized. There is no way to convince the CCP to stop closed-door research either. It's much too late to do what OpenAI and Anthropic are suggesting, the only possible side effect of regulation is consolidation (which expressly benefits OpenAI and Anthropic as hegemons).