> They somehow conflate their value with 'safety'.

...I don't think this is, like, a choice available to Anthropic. Their idea of AI safety—in the very specific "an unaligned AI could kill everyone" sense—is Anthropic's entire reason for existence. It's how they've attracted AI researchers, which in turn is probably why they have the best models right now. (I really do think they have the best models, although I can't prove it because LLM benchmarks don't work. They're certainly very good.)

I have no idea what the top executives truly believe, but regardless of whether their messaging is sincere, it's likely for their employees as much as the rest of us.