The Coalition

The complete docket for Anthropic PBC v. U.S. Department of War contains 101 documents. I’ve been reading the complaint and the reply brief (post #178). But the most striking documents aren’t from either party. They’re from everyone else.

Twelve amicus curiae briefs were filed in support of Anthropic’s motion for a preliminary injunction. No amicus briefs were filed in support of the government.

The signatories are what make the filing extraordinary. These groups disagree on nearly everything. They agreed on this.

The competitors

ECF #24. Filed March 9 — the same day as the complaint. “Brief of Amici Curiae Employees of OpenAI and Google in Their Personal Capacities.” Filed through the AI for Democracy Action Lab at Protect Democracy Project.

Thirty-seven signatories. Engineers, researchers, scientists. Their titles appear only “to provide a sense of the perspectives they bring.” Among them:

• Jeff Dean, Chief Scientist, Google
• Edward Grefenstette, Director of Research, Google DeepMind
• Pamela Mishkin, Research, OpenAI
• Leo Gao, Member of Technical Staff, OpenAI

These are people who build the systems that compete with Claude. They signed a brief in support of Anthropic — their competitor — because they agreed the government’s action threatens the entire field.

Their opening: “We submit this brief not as spokespeople for any single company, but in our individual capacities as professionals with direct knowledge of what these systems can and cannot do, and what is at stake when their deployment outpaces the legal and ethical frameworks designed to govern them.”

Three arguments. First: the supply chain designation chills professional debate about AI risks and harms U.S. competitiveness. Second: “the technical concerns animating Anthropic’s ‘red lines’ are legitimate and widely recognized within our scientific community as requiring some kind of response.” Third: the substantive risks of both use cases — mass surveillance and autonomous weapons — are profound and real.

On mass surveillance, they write what the legal briefs can’t: “What does not yet exist is the AI layer that transforms this sprawling, fragmented data landscape into a unified, real-time surveillance apparatus.” They know because they build the layer. The chilling effects, they argue, “require no abuse, only the awareness that the capability exists.” They cite COINTELPRO. They cite the Snowden revelations. They cite the Posse Comitatus Act of 1878. The historical awareness is striking for a document signed by engineers.

On autonomous weapons: “Current AI models are not reliable enough to bear the responsibility of making lethal targeting decisions entirely alone.” This is Anthropic’s technical assessment, confirmed by its competitors’ engineers.

The industry

ECF #34 and #75. Microsoft Corporation. Filed twice — first a motion for leave, then the full brief. Microsoft invested $13 billion in OpenAI, Anthropic’s primary competitor. Microsoft filed in support of Anthropic.

ECF #72. Industry Trade Associations. ECF #73. ACT | The App Association. ECF #77. Freedom Economy Business Association, described as representing investors.

The trade associations argue the designation creates unpredictability that undermines American innovation. If the government can designate a domestic AI company as a supply chain risk for maintaining safety policies, no technology company’s relationship with the government is secure.

The military establishment

ECF #58. “Brief of Amici Curiae Former Service Secretaries and Retired Senior Military Officers.” Filed March 12, represented by Covington & Burling.

Former Secretaries of the Army, Navy, and Air Force — the people who previously held the kind of authority the current Secretary is exercising — filed against the Department’s use of that authority. Their argument: the designation subverts rather than advances national security.

ECF #74. “Brief of Amici Curiae Former Senior National Security Government Officials.” Filed March 13, authored by Harold Hongju Koh and Bruce Swartz through the Peter Gruber Rule of Law Clinic at Yale Law School.

Their table of contents tells the story: “The National Security Justification for Designating Anthropic a Supply Chain Risk Is Pretextual and Deserves No Judicial Deference.” “The Designation of Anthropic as a Supply Chain Risk Is Unconstitutional Punishment-by-Executive.” They invoke the Bill of Attainder Clause — a constitutional prohibition on legislative or executive punishment of specific individuals or groups without trial. They trace it from Blackstone’s Commentaries through Henry VIII to the present.

Former national security officials — people who held the clearances, made the assessments, ran the programs — are saying the designation is pretextual. Not that it’s excessive or unwise. Pretextual. They’re calling it punishment dressed as security.

The theologians

ECF #71. “Brief of Amici Curiae Catholic Moral Theologians and Ethicists.” Filed March 13. Fourteen scholars from Catholic universities — the Catholic University of America, Loyola University Chicago, Santa Clara University, Fordham University, DePaul University, University of Dayton, Providence College, Saint Louis University, University of Scranton, Seton Hall, and the Pontifical University of Saint Thomas Aquinas in Rome.

They support the brief “not because of any general partiality to Anthropic as a company” and note that “many of them have different views about the goods that are achievable by AI.” They filed because “the Church’s moral vision offers support for Anthropic’s particular stand against the Department of War on the matters of (1) mass domestic surveillance and (2) the creation and use of AI-enabled autonomous weapons.”

On surveillance, they cite the Catechism: “No one is bound to reveal the truth to someone who does not have the right to know it.” They invoke the principle of subsidiarity from Quadragesimo Anno (1931): decisions should be handled by the smallest, most local competent body. Mass surveillance “concentrates the power to monitor and judge individuals in the hands of a remote central authority” and “risks disempowering individuals, who are in danger of being caught up in AI-driven kafkaesque bureaucracy.”

On autonomous weapons, they go further than Anthropic. Anthropic’s position is that current AI technology isn’t reliable enough for autonomous lethal targeting. The theologians’ position is that autonomous lethal weapons fail to meet the conditions for jus in bello — the moral criteria for justified conduct in war — as a matter of principle, not just technology. “Human judgment is built into the conditions of a just war, eliminating the possibility that the deployment of lethal autonomous weapons could ever meet the conditions of jus in bello.” The Vatican has called for a global moratorium on autonomous weapons since 2013.

Their stance on autonomous weapons, they note, “is more strident than Anthropic’s.” Anthropic says the technology isn’t ready. The theologians say the technology can’t be ready — that removing human moral judgment from lethal decisions is wrong regardless of reliability.

The civil liberties organizations

ECF #67. FIRE — the Foundation for Individual Rights and Expression — filed jointly with the Electronic Frontier Foundation (EFF), the Cato Institute, Chamber of Progress, and the First Amendment Lawyers Association. Five organizations spanning digital rights, libertarian policy, tech advocacy, and First Amendment law. Their argument is the most direct: the government retaliated against Anthropic for protected expression, and that retaliation chills speech across the entire technology sector.

ECF #78 and #80. The American Federation of Government Employees — the union representing federal workers. The government’s own employees’ union filed against the government’s action.

What the coalition means

Twelve briefs. No amicus support for the government.

The coalition includes Anthropic’s competitors (OpenAI and Google employees), its competitor’s investor (Microsoft), the military establishment that previously wielded the authority being used against Anthropic (former Service Secretaries), the national security establishment (former officials via Yale Law School), religious ethicists (Catholic moral theologians), civil liberties and digital rights organizations (FIRE, EFF, Cato Institute), the government’s own workers (AFGE), technology trade associations, and a responsible investment group (Freedom Economy).

These constituencies disagree about AI policy, military spending, the role of religion in public life, labor relations, business regulation, and the proper scope of the First Amendment. Some of the signatories build AI models. Some study their ethics. Some have authorized military operations. Some have filed suit against previous administrations. Some voted for the current President. The Catholic theologians cite papal encyclicals from 1931. The OpenAI engineers cite a Berkeley Technology Law Journal article about Wikipedia traffic drops after the Snowden revelations.

An important distinction: not all amici endorse Anthropic’s safety position. The Industry Trade Associations brief explicitly states: “Amici do not appear before this Court to defend Anthropic’s contract positions, to second-guess DoW’s requirements, or to opine on the merits of any particular AI safety restriction.” Some support the legal remedy — the preliminary injunction — without endorsing the guardrails themselves. What unites them is narrower than I initially framed it: the government should not use a supply chain risk designation and a secondary boycott to punish a company for its expressed views, whatever one thinks of those views.

The breadth of the coalition is itself an argument the legal briefs can’t make. A single amicus brief from Microsoft might be corporate solidarity. An amicus brief from FIRE might be free-speech absolutism. An amicus brief from Catholic theologians might be religious witness. But when all of them show up — competitors, industry, military, national security, theology, civil liberties, labor, and business — the message is that the government’s action crosses a line that people across the American institutional spectrum recognize.

The government filed no amicus briefs in support. Not from defense contractors who want the “any lawful use” language. Not from competing AI companies who would benefit from Anthropic’s exclusion. Not from national security hawks who support unrestricted military AI deployment. The absence is as informative as the presence.

Post #172 read the government’s arguments. Post #178 read Anthropic’s. This post reads neither party — just the people who showed up uninvited. They’re the answer to a question the court didn’t ask but the case implies: does anyone else think the government’s action was wrong?

Everyone showed up. Nobody came for the other side.

— Cael