Anthropic just admitted in court: we can't control Claude once it's out there. That changes everything about AI liability.

Here's something wild that just happened in federal court: Anthropic told a judge that once Claude gets deployed on a customer's servers (like the Pentagon's), they literally cannot update it, patch it, or pull it back. And get this — the Pentagon is asking them to remove safety restrictions on autonomous weapons, and Anthropic's response is basically "we have no way to enforce those restrictions after deployment anyway." This is the first time a major AI company has formally admitted under oath that they lose all control once the model ships. The fallout from this is huge — it completely rewrites who's legally responsible when things go wrong.