Anthropic wins US court fight over Pentagon AI blacklist

The California ruling that the Pentagon’s designation of Anthropic as a supply chain risk was “illegal and baseless” is a significant moment in how democratic systems will referee power struggles between AI labs and security agencies. For months, the Trump administration had tried to force Anthropic to drop its bans on mass surveillance of Americans and fully autonomous weapons, using procurement tools and blacklist style labels as leverage. The court’s message is that you cannot punish a vendor’s protected speech and safety stance by quietly cutting it out of the broader federal market.

For the race to AGI, the decision removes a cloud that hung over one of the three leading frontier labs. Anthropic still faces IP lawsuits and political pushback, but it no longer carries an official “supply chain risk” brand that could have scared off commercial partners or foreign governments. That stabilizes its ability to sign long term deals and finance ever larger models, including the Mythos family. At the same time, the ruling does not resolve the underlying policy question of who decides acceptable military uses of general purpose models. Other agencies can experiment with different legal tools, and future administrations may seek new authorities. The episode previews a world where AGI scale vendors will litigate against their own governments when safety policies collide with national security demands.