In my experience Anthropic positions itself as the "safe" AI company more than the "ethical" AI company. They're related but not the same thing.

The only way you could be surprised that Anthropic wants to be in bed with the US military is if you just never listened to anything Dario has said publicly. He's very open about wanting the US government and the US military to use Claude to win against China. That's why Claude was in the Pentagon before all the others in the first place.

>LLMs are fundamentally incapable of controlling autonomous weapons anyway

This is obviously false, though that's not surprising from what I've seen from Zitron. Claude is probably too slow and clunky to go full mech warrior for the time being, but it would be trivial to hook Claude up to an autonomous drone with missile strike capabilities. Those things are mostly autonomous already, they just require a human to tell them where to shoot. Claude can easily do that with a simple API.

The rest is valid. I wouldn't describe Anthropic as an ethical company. On the contrary, if you believe that you losing the AI race is an existential threat to humanity, then it's easy to justify all sorts of unethical behavior for the greater good.