Center for Democracy & Technology 84d ago

On March 16, the federal district court for the Northern District of California issued a preliminary injunction (PI) against the Department of Defense (Department) for designating Anthropic a “supply-chain risk.” The designation was in retaliation against the company for insisting on contractual language that would bar the use of its AI model for mass surveillance of Americans and in autonomous lethal weapons. The court found that the designation “likely violated [Anthropic’s] due process rights” and that the Department had “flouted procedural safeguards required by Congress” for such a designation. It also enjoined an alphabet soup list of other agencies — ranging from the State Department to Health and Human Services — from implementing the February 27 Presidential Directive that ordered them to cease doing business with Anthropic. The government has appealed the PI to the Ninth Circuit but has not to this point sought a stay, so the PI has now gone into effect. The Department also issued a designation under a separate statute. Anthropic has challenged that designation in the D.C. Circuit (the venue mandated by that statute), and the court has yet to rule in that case, in which CDT and the American Civil Liberties Union filed an amicus brief in support of Anthropic. Our brief highlights the importance of the mass surveillance issues Anthropic was punished for raising in the public debate. We are hopeful that the D.C. Circuit will also strike down this coercive retaliatory action by the Department. Although the dispute in the two cases specifically concerns Anthropic, it raises issues of broad concern. A few days prior to the California court decision, CDT hosted a Working Groups call to discuss these issues. The speakers on the call, Paul Lekas of the Software & Information Industry Association, Jennifer Huddleston of the Cato Institute, Alan Rozenshtein of the University of Minnesota Law School and CDT’s Samir Jain, Jake Laperruque and Quinn Anex-Ries called into question the lawfulness and constitutionality of the designation, decried its impact on industry and governmental agencies, and highlighted the danger to privacy of AI-powered mass surveillance of Americans. Normally, CDT Working Groups calls, which are by invitation only, are conducted under the Chatham House rule. While we enforced that rule for the Q&A portion of this call, we are making public the speakers’ affirmative presentations (with permission) because of the importance of the public discourse surrounding the guardrails that Anthropic is seeking to maintain on the Department’s use of Anthropic’s AI systems.