1. United States
2. N.Y.
3. Letter

The Trump administration's treatment of Anthropic is a constitutional crisis dressed up as AI policy, and I want you to push back on it. After Anthropic refused Pentagon demands to use its models for autonomous weapons development and surveillance of Americans, the government declared the company a "supply chain risk" — effectively banning federal agencies and contractors from working with them. A court had to step in to block sanctions that would have cost Anthropic hundreds of millions of dollars. That's not regulation. That's retaliation. Now a new executive order has imposed export controls specifically targeting Anthropic's Mythos and Fable models, forcing the company to shut them down entirely. Other AI models with comparable offensive cybersecurity capabilities face no such restrictions — only a voluntary framework. The EFF's amicus brief in Anthropic v. Department of War makes the First Amendment case clearly: code is speech, and punishing a company for its public refusal to build weapons is unconstitutional. Cybersecurity experts at freefable.org warn these sanctions actually weaken our defenses by blocking developers from using the best tools to find vulnerabilities before adversaries do. Real risks may justify real rules — but those rules must be rational and applied evenhandedly. Right now they are neither. Demand oversight hearings on how these export controls were justified, and make clear that procurement and trade policy cannot be weaponized against companies that refuse government demands.