The key difference is that Anthropic aired their disagreement with the DoD publicly, and the DoD is not going to work with a company that tries to exert any amount of control over their relationship via the public sphere. Same goes for Trump.

I think Anthropic knew full well that by publishing their disagreement, it would sink the deal and relationship, and I think they also calculated (correctly) that that act of defiance would get them good publicity and potentially peel away some of OpenAIs user base. I think this profit incentive happened to align with their morals, and now here we are.