LawAI's Charlie Bullock on why AI labs keep two safety policies, the real one, and the one they're legally bound to: "The way Anthropic complies with this is they have their responsible scaling policy that their safety team works on, and that's very robust. But that's not the one that counts for SB 53. The one that counts is a separate document they made for compliance reasons that doesn't bind them to do anything." "Because if you commit to something and then you're legally obligated to…