Overview
- Judges on the D.C. Circuit pressed both sides during Tuesday's hearing, with one judge calling the Defense Department's move a potential overreach and another stressing the unpredictability of fast-changing AI models.
- The Pentagon labeled Anthropic a supply chain risk in March after talks collapsed over an “all lawful use” demand, a tag that forces defense vendors to certify they will not use Anthropic’s Claude models in military work.
- The Justice Department argued Anthropic could bake limits into future models and cannot be trusted, while Anthropic said it cannot alter tools once deployed on classified networks and called the designation unlawful and punitive.
- The blacklist remains in effect after the D.C. Circuit denied a temporary block in April yet expedited the case, and a separate San Francisco injunction lets non-Pentagon agencies keep using Anthropic’s systems during the fight.
- The panel took the appeal under advisement with a decision expected in weeks, and President Trump has ordered agencies to remove Anthropic within six months, raising stakes for defense contractors who must juggle compliance and capability.