Anthropic PBC v. U.S. Department of War

Analysis of the Original Complaint, the Clash Between AI Ethics and National Security

March 19, 2026 | KIMKJ.COM Analysis

Case Overview

On March 9, 2026, AI company Anthropic PBCfiled two federal lawsuits simultaneously against the U.S. Department of War and Secretary of Defense Pete Hegseth. One was filed in the U.S. District Court for the Northern District of California (Case No. 3:26-cv-01996), and the other was filed in the U.S. Court of Appeals for the D.C. Circuit (Case No. 26-1049).

The core of this lawsuit is the illegality of the Trump administration's administrative action designating Anthropic as a 'Supply-Chain Risk to National Security'. This designation has usually been applied to foreign adversarial forces, such as cybersecurity companies with ties to the Chinese military. Its use against a U.S. domestic company is unprecedented in modern history.

The complaint was filed by WilmerHale(Wilmer Cutler Pickering Hale and Dorr LLP), with attorneys Michael J. Mongan of San Francisco, Kelly P. Dunbar, Joshua A. Geltzer, Kevin M. Lamb, Anneke Dunbar-Gronke of Washington, D.C., and Emily Barnet of New York listed as counsel.

Background of the Dispute, Two 'Red Lines'

Since 2024, Anthropic has supplied Claude AI to the Department of Defense through military contractor Palantir, and in July 2025 it entered into a direct Department of Defense contract worth $200 million. Anthropic is also the first frontier AI lab approved for access to U.S. government classified networks. Claude was used for lawful purposes such as data analysis and support for intelligence assessment.

But during the contract renewal process, Anthropic maintained two usage restrictions, the so-called red lines.

• Ban on mass surveillance of U.S. citizens - Claude must not be used for mass surveillance of U.S. citizens.

• Ban on deployment of fully autonomous lethal weapons - Claude must not be used in weapons systems that make lethal decisions using AI alone, without human target selection and firing decisions.

The complaint states that these restrictions are directly tied to Anthropic's mission as a Public Benefit Corporation. Allowing Claude to be used for mass surveillance and lethal weapons without human oversight would be incompatible with Anthropic's founding principles.

In response, Secretary of Defense Hegseth argued that the Department of Defense must be able to use AI for any lawful purpose, and that private companies should not exercise veto power over military operations.

Facts Alleged in the Complaint (Timeline)

• 2024 - Anthropic begins supplying Claude to the Department of Defense through Palantir

• July 2025 - Direct $200 million contract signed with the Department of Defense

• February 2026 - CEO Dario Amodei publicly declares the two red lines and meets with Secretary Hegseth.

• February 27, 2026 - Secretary Hegseth posts a public directive on X, formerly Twitter, ordering Anthropic to be designated as a 'supply-chain national security risk.' President Trump instructs all federal agencies to stop using Anthropic technology.

• March 3, 2026 - Department of Defense issues an official Notice under 41 U.S.C. § 4713. Effective immediately.

• March 4, 2026 - Notice reaches Anthropic. GSA removes Anthropic from USAspend.gov.

• March 9, 2026 - Anthropic files two federal lawsuits

Secretary Hegseth's X Post (Attachment 1)

Addendum 1 to the complaint includes the full text posted by Secretary Hegseth on his official X account, @SecWar, at 5:14 p.m. ET on February 27, 2026. This post has the nature of a public directive designating Anthropic as a supply-chain risk and was cited in the complaint as key evidence.

Legal Causes of Action in the Complaint (Five Counts)

Anthropic's complaint runs 78 pages and presents five causes of action, or counts.

Count I — Violation of the Administrative Procedure Act (APA): The Department of Defense failed to follow the procedures required by Congress when designating a supply-chain risk under the Federal Acquisition Supply Chain Security Act (FASCSA, 41 U.S.C. § 4713). The complaint argues that the Department's action was arbitrary, capricious, and an abuse of discretion.

Count II — Violation of the First Amendment (retaliation for free speech): The government imposed economic retaliation because of Anthropic's protected speech on AI safety. It was punishment for Anthropic publicly expressing views on the limits of AI services and on the public issue of AI safety.

Count III — Violation of the First Amendment (viewpoint discrimination): The government took discriminatory action because of a specific viewpoint, Anthropic's position on AI safety. The key issue is differential treatment compared with competitors such as OpenAI and xAI, which accessed government networks without similar restrictions.

Count IV — Violation of the Fifth Amendment (due process): The supply-chain risk designation was made without proper procedural safeguards, including notice, a hearing, and an opportunity to respond.

Count V — APA violation (unilateral contract termination): The federal agencies' unilateral cessation of Anthropic technology use pursuant to President Trump's instruction did not go through lawful procedures under the Administrative Procedure Act.

D.C. Circuit Petition for Review

The petition filed in the U.S. Court of Appeals for the D.C. Circuit, 13 pages, is based on 41 U.S.C. § 1327(b) and Federal Rule of Appellate Procedure 15(a). This court has direct judicial review jurisdiction over supply-chain risk determinations under FASCSA, 41 U.S.C. § 4713.

The petition's central arguments are that the Department of Defense's action was (1) a pretextual form of retaliation in violation of the First and Fifth Amendments, (2) arbitrary, capricious, and an abuse of discretion, (3) unsupported by the administrative record, (4) inconsistent with the procedures required by law, and (5) beyond statutory authority.

The Addendum attached to the petition includes Secretary Hegseth's X post (Addendum 1a) and the official notice under 41 U.S.C. § 4713 (Addendum 2a-5a). According to the notice, this designation is 'effective immediately' and remains valid until modified or terminated in writing by the Secretary.

Key Quotes from the Complaint

Legal Grounds and Jurisdiction

Anthropic filed suits in separate courts based on two different legal frameworks.

• U.S. Court of Appeals for the D.C. Circuit (Case No. 26-1049) - Petition for judicial review of the supply-chain risk determination under the Federal Acquisition Supply Chain Security Act (FASCSA, 41 U.S.C. § 4713). This statute grants direct review jurisdiction to the D.C. Circuit (41 U.S.C. § 1327(b)(3)).

• U.S. District Court for the Northern District of California (Case No. 3:26-cv-01996) - Challenge to the separate supply-chain risk designation authority invoked by Secretary Hegseth under 10 U.S.C. § 3252. Judge Rita F. Lin was assigned to this court.

According to legal expert Zach Prince, a partner at Haynes Boone, 10 U.S.C. § 3252 narrowly defines covered systems as national security systems in the information or weapons fields, so applying this law to government-wide procurement may have exceeded legal authority.

Prayer for Relief

Anthropic asked the court for the following relief.

• Temporary restraining order (TRO) and preliminary injunction - Immediate suspension of enforcement of the supply-chain risk designation while the litigation proceeds

• Declaratory judgment - Declaration that the supply-chain risk designation is unlawful and unconstitutional

• Permanent injunction - Prohibition against the government permanently enforcing this designation

Contradictions in the Government's Arguments

There is a central contradiction in the government's position, as shown in the complaint and related reporting. The Department of Defense claimed that Anthropic was a risk to national security, while at the same time allowing a six-month transition period for continued service. If it were a genuine security risk, immediate blocking would be necessary, and the six-month grace period suggests that the designation is political retaliation, not a security risk. Reports also indicate that the Department of Defense considered invoking the Defense Production Act to argue that Claude is 'essential' to national security, which is logically hard to reconcile with the simultaneous designation of it as a 'supply-chain risk.'

Status of Amicus Brief Filings

This lawsuit has received unusually broad support. By March 17, roughly 150 current and former federal and state judges had filed amicus briefs supporting Anthropic. More than 30 employees of OpenAI and Google DeepMind also released statements supporting Anthropic. Major technology industry groups, including SIIA, CCIA, ITI, and TechNet, filed a joint amicus brief, and the Cato Institute, the Society for the Rule of Law, and the Chamber of Progress also joined.

Anthropic's Corporate Disclosure

According to the Rule 26.1 disclosure statement in the D.C. Circuit petition, Anthropic PBC is a privately held public benefit corporation and has no parent company. There are two publicly traded companies that hold 10% or more of its shares. (1) Google LLC - a wholly owned subsidiary of Alphabet, Inc., and (2) Amazon Web Services, Inc. - a wholly owned subsidiary of Amazon.com, Inc. This disclosure officially confirms Anthropic's investor structure and shows that both Google and Amazon, two giant companies, hold stakes of 10% or more.

Significance of This Case

This lawsuit is the first case to legally draw the boundary between corporate ethical autonomy and national security demands in the AI era. Depending on the ruling, it will determine whether AI companies can have a say in usage conditions in government contracts, and whether the government can impose economic sanctions because of a company's policy position. Attorney Prince of Haynes Boone predicted that Anthropic is likely to obtain an injunction, but added that national security arguments may delay early court intervention. The injunction hearing is scheduled for Tuesday, March 25.