As far as I can tell, and I'm sure someone will correct me if I'm wrong, the OpenAI version was, we'll let the government do whatever it wants, that's specified in law. Whereas Anthropic seemed to say, we're not going to cross those lines, even if they are permitted by law.
Important distinction as there are now no laws against autonomous weapons and surveillance has been widespread and unhindered by judicial oversight for quite a while now.