EFF: Trump Admin Retaliated Against Anthropic for Refusing Autonomous Weapons Work

The Electronic Frontier Foundation (EFF) has filed an amicus brief arguing that the Pentagon's sanctions against Anthropic violate the First Amendment, as they were motivated by retaliation—not genuine national security concerns. The government designated Anthropic a “supply chain risk” after the company refused to let its AI models be used for fully autonomous killing or mass surveillance of Americans. A court has issued a preliminary injunction blocking the sanctions, which would have cost the company hundreds of millions of dollars.

Key Events

• Retaliation Trigger: Anthropic resisted government demands to use its models for autonomous weapons or spying on Americans. The government then declared the company a “supply chain risk,” effectively banning federal agencies and contractors from doing business with it.
• Export Controls on Mythos/Fable: A recent executive order imposed “export controls” on Anthropic's new Mythos and Fable models, banning foreign nationals from using them. Anthropic shut down the models entirely to comply. The administration claimed Mythos could find vulnerabilities in code—a capability EFF notes is not unique among LLMs.
• Disparate Treatment: Other LLMs with similar offensive cybersecurity capabilities face only a voluntary 30-day pre-release testing framework, while Anthropic gets punitive export controls.

EFF's Argument

EFF and allies argue that the sanctions are unconstitutional retaliation for Anthropic's public refusal to enable autonomous weapons. They call for AI policy that is “reasonably responsive to real-world risk, grounded in the realities of the technology, and no more burdensome than necessary.” The brief urges cutting through hype rather than feeding it, noting that even if Mythos's capabilities are modest improvements, competitors are closing the gap.