ANTHOPIC PBC v. U.S. DEPARTMENT OF WAR, COMPLANT FOR DECLARATORY AND INJUNTIVE RELIEF

https://en.wikipedia.org/wiki/Anthropic

ANTHOPIF PBC v. U.S. DEPARTMENT OF WAR, COMPLANT FOR DECLARATORY AND INJUNCTIVE RELIEF 

cv-01996-RFL Document 1 Filed 03/09/26 Page 1 of 48

MICHAEL J. MONGAN (SBN 250374)

michael.mongan@wilmerhale.com

WILMER CUTLER PICKERING

HALE AND DORR LLP

50 California Street, Suite 3600

San Francisco, CA 94111

Telephone: (628) 235-1000

EMILY BARNET (pro hac forthcoming)

emily.barnet@wilmerhale.com

WILMER CUTLER PICKERING

HALE AND DORR LLP

7 World Trade Center

250 Greenwich St

New York, NY 10007

Telephone: (212) 230-8800

Attorneys for Plaintiff Anthropic PBC

KELLY P. DUNBAR (pro hac forthcoming)

kelly.dunbar@wilmerhale.com

JOSHUA A. GELTZER (pro hac forthcoming)

joshua.geltzer@wilmerhale.com

KEVIN M. LAMB (pro hac forthcoming)

kevin.lamb@wilmerhale.com

SUSAN HENNESSEY (pro hac forthcoming)

susan.hennessey@wilmerhale.com

LAUREN MOXLEY BEATTY (SBN 308333)

(application pending)

lauren.beatty@wilmerhale.com

LAURA E. POWELL (pro hac forthcoming)

laura.powell@wilmerhale.com

SONIKA R. DATA (pro hac forthcoming)

sonika.data@wilmerhale.com

WILMER CUTLER PICKERING

HALE AND DORR LLP

2100 Pennsylvania Avenue NW

Washington, DC 20037

Telephone: (202) 663-6000

UNITED STATES DISTRICT COURT

NORTHERN DISTRICT OF CALIFORNIA

ANTHROPIC PBC,

Plaintiff,

v.

U.S. DEPARTMENT OF WAR,

U.S. DEPARTMENT OF THE TREASURY,

FEDERAL HOUSING FINANCE AGENCY,

U.S. DEPARTMENT OF STATE,

COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF

U.S. DEPARTMENT OF HEALTH AND

HUMAN SERVICES,

U.S. DEPARTMENT OF COMMERCE,

U.S. DEPARTMENT OF VETERANS

AFFAIRS,

GENERAL SERVICES

ADMINISTRATION,

COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 2 of 48

U.S. OFFICE OF PERSONNEL

MANAGEMENT,

U.S. NUCLEAR REGULATORY

COMMISSION,

U.S. SOCIAL SECURITY

ADMINISTRATION,

U.S. DEPARTMENT OF HOMELAND

SECURITY,

SECURITIES AND EXCHANGE

COMMISSION,

NATIONAL AERONAUTICS AND SPACE

ADMINISTRATION,

U.S. DEPARTMENT OF ENERGY,

FEDERAL RESERVE BOARD OF

GOVERNORS,

NATIONAL ENDOWMENT FOR THE

ARTS,

EXECUTIVE OFFICE OF THE

PRESIDENT,

PETER B. HEGSETH, in his official capacity

as Secretary of War,

SCOTT BESSENT, in his official capacity as

Secretary of the Treasury,

WILLIAM J. PULTE, in his official capacity

as Director of U.S. Federal Housing,

MARCO RUBIO, in his official capacity as

Secretary of State,

ROBERT F. KENNEDY, JR., in his official

capacity as Secretary of Health and Human

Services,

COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 3 of 48

HOWARD LUTNICK, in his official capacity

as Secretary of Commerce,

DOUGLAS A. COLLINS, in his official

capacity as Secretary of Veterans Affairs,

EDWARD C. FORST, in his official capacity

as Administrator of the General Services

Administration,

SCOTT KUPOR, in his official capacity as

Director of the U.S. Office of Personnel

Management,

HO K. NIEH, in his official capacity as

Chairman of the U.S. Nuclear Regulatory

Commission,

FRANK J. BISIGANO, in his official

capacity as Commissioner of the U.S. Social

Security Administration,

KRISTI NOEM, in her official capacity as

Secretary of Homeland Security,

PAUL S. ATKINS, in his official capacity as

Chairman of the Securities and Exchange

Commission,

JARED ISAACMAN, in his official capacity

as Administrator of the National Aeronautics

and Space Administration,

CHRIS WRIGHT, in his official capacity as

Secretary of Energy,

JEROME H. POWELL, in his official

capacity as Chairman of the Federal Reserve

Board of Governors,

MARY ANNE CARTER, in her official

capacity as Chairman of the National

Endowment for the Arts, and

DOE DEFENDANTS 1-10,

Defendants.

COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 4 of 48

INTRODUCTION

1. Anthropic is a leading frontier artificial intelligence (AI) developer whose

flagship family of AI models is known as “Claude.” Anthropic was founded based on the belief

that AI technologies should be developed and used in a way that maximizes positive outcomes

for humanity, and its primary animating principle is that the most capable artificial-intelligence

systems should also be the safest and the most responsible. Anthropic brings this suit because the

federal government has retaliated against it for expressing that principle. When Anthropic held

fast to its judgment that Claude cannot safely or reliably be used for autonomous lethal warfare

and mass surveillance of Americans, the President directed every federal agency to

“IMMEDIATELY CEASE all use of Anthropic’s technology”—even though the Department of

War (Department) had previously agreed to those same conditions. Hours later, the Secretary of

War directed his Department to designate Anthropic a “Supply-Chain Risk to National Security,”

and further directed that “effective immediately, no contractor, supplier, or partner that does

business with the United States military may conduct any commercial activity with Anthropic.”

In a letter to Anthropic, the Secretary confirmed the designation as “necessary to protect national

security.”1 These actions are unprecedented and unlawful. The Constitution does not allow the

government to wield its enormous power to punish a company for its protected speech. No

federal statute authorizes the actions taken here. Anthropic turns to the judiciary as a last resort

to vindicate its rights and halt the Executive’s unlawful campaign of retaliation.

2. Since its inception, Anthropic has worked to offer AI services to customers in the

private and public sectors in a manner consistent with its founding principles of safety and

responsibility. It has partnered extensively with the federal government, and particularly the

United States Department of War. Anthropic has even developed Claude models that help the

Department to protect national security. As a result of these efforts, Claude is reportedly the

Department’s most widely deployed and used frontier AI model, and the only frontier AI model

on the Department’s classified systems. And the Department has acknowledged Anthropic’s

1 See Exhibit 3 (March 3, 2026 Letter).

- 1 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 5 of 48

unique contributions in this area, praising Claude for its “exquisite” capabilities and reportedly

using Claude—to this day—in its most important military missions.

3. Anthropic’s Usage Policy has always conveyed its view that Claude should not be

used for two specific applications: (1) lethal autonomous warfare and (2) surveillance of

Americans en masse. Anthropic has never tested Claude for those uses. Anthropic currently does

not have confidence, for example, that Claude would function reliably or safely if used to

support lethal autonomous warfare. These usage restrictions are therefore rooted in Anthropic’s

unique understanding of Claude’s risks and limitations—including Claude’s capacity to make

mistakes and its unprecedented ability to accelerate and automate analysis of massive amounts of

data, including data about American citizens. Anthropic has collaborated with the Department of

War on modifications to its usage restrictions to facilitate the Department’s work with Claude, in

recognition of the Department’s unique missions. But Anthropic has always maintained its

commitment to those two specific restrictions, including in its work with the Department of War.

4. Recently, however, Secretary of War Hegseth and his Department began

demanding that Anthropic discard its usage restrictions altogether and replace them with a

general policy under which the Department may make “all lawful use” of the technology.

Anthropic largely agreed to the Department’s request, except for two restrictions it viewed as

critical: prohibitions against use of the technology for lethal autonomous warfare and mass

surveillance of Americans. Throughout these discussions, Anthropic expressed its strongly held

views about the limitations of its AI services. It also made clear that, if an arrangement

acceptable to the Department could not be reached, Anthropic would collaborate with the

Department on an orderly transition to another AI provider willing to meet its demands.

5. The Department met Anthropic’s attempts at compromise with public castigation.

It labeled Anthropic’s CEO as too “ideological” and a “liar” with a “God-complex” who “is ok

putting our nation’s safety at risk.” The Department eventually gave Anthropic a public

ultimatum: “get on board” and accede to the government’s demands by 5:01 p.m. on February

27, 2026, or “pay a price” in the form of either being cast out of the defense supply chain under

- 2 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 6 of 48

10 U.S.C. § 3252 or forced to provide unlimited use of Claude under the Defense Production

Act.

6. After Anthropic’s CEO publicly announced that the company could not “in good

conscience accede to” the Department’s demands, the Executive Branch swiftly retaliated.

7. On February 27, 2026, President Trump posted a statement on social media (the

Presidential Directive), “directing EVERY Federal Agency in the United States Government to

IMMEDIATELY CEASE all use of Anthropic’s technology.” He derided Anthropic as “out-of-

control” and a “RADICAL LEFT, WOKE COMPANY” of “Leftwing nut jobs.” He also accused

Anthropic of “selfishness” and of making a “DISASTROUS MISTAKE.” “Anthropic better get

their act together,” the President threatened, or he would “use the Full Power of the Presidency to

make them comply, with major civil and criminal consequences to follow.”

8. The same afternoon, Secretary Hegseth purported to act on “the President’s

directive” by posting a “final” decision via social media (the Secretarial Order). The Secretarial

Order “direct[ed] the Department of War to designate Anthropic a Supply-Chain Risk to

National Security.” It also proclaimed that “[e]ffective immediately, no contractor, supplier, or

partner that does business with the United States military may conduct any commercial activity

with Anthropic.” The Secretary denounced what he characterized as Anthropic’s “Silicon Valley

ideology,” “defective altruism,” “corporate virtue-signaling,” and “master class in arrogance.”

And he criticized Anthropic for not being “more patriotic.” But he also directed that “Anthropic

will continue to provide the Department of War its services for a period of no more than six

months.”

9. Other federal agencies soon followed suit. For example, the General Services

Administration terminated Anthropic’s “OneGov” contract, thereby ending the availability of

Anthropic services to all three branches of the federal government. The Department of the

Treasury and the Federal Housing Finance Agency publicly stated they were cutting ties with

Anthropic. And the Departments of State and Health and Human Services reportedly circulated

internal memoranda directing employees to stop using Anthropic’s services.

- 3 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 7 of 48

10. On March 4, 2026, at 8:48 p.m. Eastern, the Secretary of War sent Anthropic a

letter about the “supply chain risk” designation in the Secretarial Order. That letter (the

Secretarial Letter), dated March 3, notified Anthropic that “the Department of War (DoW) has

determined . . . that the use of [Anthropic’s] products in [the Department’s] covered systems

presents a supply chain risk” and that exercising the authority granted by 10 U.S.C. § 3252

against Anthropic is “necessary to protect national security.” The Secretarial Letter pronounces

that this determination covers all Anthropic “products” and “services,

” including any that

“become available for procurement.

” And it asserts that “less intrusive measures are not

reasonably available” to mitigate the risks that Anthropic’s products and services supposedly

pose to national security.

11. All of these unprecedented actions—the Presidential Directive, the Secretarial

Order and the Secretarial Letter that followed it, and other agency actions taken in response to

the Presidential Directive (collectively, the Challenged Actions)—are harming Anthropic

irreparably. In Secretary Hegseth’s own words, Anthropic’s status in the eyes of the federal

government has been “permanently altered.” Official designation as a “Supply-Chain Risk to

National Security” carries profound weight, particularly under a President who has threatened

both “criminal consequences” and “the Full Power of the Presidency” to enforce compliance.

Anthropic’s contracts with the federal government are already being canceled. Current and future

contracts with private parties are also in doubt, jeopardizing hundreds of millions of dollars in

the near-term. On top of those immediate economic harms, Anthropic’s reputation and core First

Amendment freedoms are under attack. Absent judicial relief, those harms will only compound

in the weeks and months ahead.

12. The Challenged Actions are as unlawful as they are unprecedented. First, the

Secretarial Order “designat[ing] Anthropic a Supply-Chain Risk to National Security” and

prohibiting the Department’s contractors, suppliers, and partners from “conduct[ing] any

commercial activity with Anthropic”—and the Secretarial Letter purporting to implement the

Order—violates both 10 U.S.C. § 3252 and the Administrative Procedure Act. The Secretary’s

actions are contrary to Section 3252’s plain text, were issued without observance of the

- 4 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 8 of 48

procedures Congress required, and are arbitrary, capricious, and an abuse of discretion. Indeed,

Anthropic had been one of the government’s most trusted partners until its views clashed with

the Department’s.

13. Second, the Challenged Actions retaliated against Anthropic for its speech and

other protected activities in violation of the First Amendment. The Constitution confers on

Anthropic the right to express its views—both publicly and to the government—about the

limitations of its own AI services and important issues of AI safety. The government does not

have to agree with those views. Nor does it have to use Anthropic’s products. But the

government may not employ “the power of the State to punish or suppress [Anthropic’s]

disfavored expression.” Nat’l Rifle Ass’n of Am. v. Vullo, 602 U.S. 175, 188 (2024).

14. Third, the Presidential Directive requiring every federal agency to immediately

cease all use of Anthropic’s technology, and actions taken by other defendants in response to that

directive, are outside any authority that Congress has granted the Executive. And “[w]hen an

executive acts ultra vires, courts are normally available to reestablish the limits on his authority.”

Chamber of Com. of U.S. v. Reich, 74 F.3d 1322, 1328 (D.C. Cir. 1996).

15. Fourth, the Challenged Actions violate the Fifth Amendment’s Due Process

Clause. Anthropic has weighty property and liberty interests in its reputation, its business

relationships, its future business prospects, and its advocacy. The Challenged Actions arbitrarily

deprive Anthropic of those interests without any process, much less due process.

16. Fifth, the Challenged Actions violate the APA’s prohibition against imposing any

“sanction,” “penalty,” “revocation,” “suspension,” or other “compulsory or restrictive” action

against a person “except within jurisdiction delegated to the agency and as authorized by law.”

5 U.S.C. §§ 551, 558.

17. The consequences of this case are enormous. The federal government retaliated

against a leading frontier AI developer for adhering to its protected viewpoint on a subject of

great public significance—AI safety and the limitations of its own AI models—in violation of

the Constitution and laws of the United States. Defendants are seeking to destroy the economic

value created by one of the world’s fastest-growing private companies, which is a leader in

- 5 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 9 of 48

responsibly developing an emergent technology of vital significance to our Nation. The

Challenged Actions inflict immediate and irreparable harm on Anthropic; on others whose

speech will be chilled; on those benefiting from the economic value the company can continue to

create; and on a global public that deserves robust dialogue and debate on what AI means for

warfare and surveillance. There is no valid justification for the Challenged Actions. The Court

should declare them unlawful and enjoin Defendants from taking any steps to implement them.

PARTIES

18. Plaintiff Anthropic is a public benefit corporation organized under the laws of

Delaware and headquartered in San Francisco. Anthropic’s customers range from Fortune 500

companies and U.S. government agencies to small businesses and individual consumers who

have integrated Claude into the core of how they work, transforming workflows on a wide range

of tasks.

19. The U.S. Department of War is a federal agency headquartered in Washington,

D.C.

20. Washington, D.C.

21. Washington, D.C.

22. The U.S. Department of the Treasury is a federal agency headquartered in

The Federal Housing Finance Agency is a federal agency headquartered in

The U.S. Department of State is a federal agency headquartered in Washington,

D.C.

23. headquartered in Washington, D.C.

24. Washington, D.C.

25. Washington, D.C.

26. Washington, D.C.

The U.S. Department of Health and Human Services is a federal agency

The U.S. Department of Commerce is a federal agency headquartered in

The U.S. Department of Veterans Affairs is a federal agency headquartered in

The General Services Administration is a federal agency headquartered in

- 6 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 10 of 48

27. Washington, D.C.

28. Rockville, Maryland.

29. Baltimore, Maryland.

30. Washington, D.C.

31. Washington, D.C.

32. headquartered in Washington, D.C.

33. The U.S. Office of Personnel Management is a federal agency headquartered in

The U.S. Nuclear Regulatory Commission is a federal agency headquartered in

The U.S. Social Security Administration is a federal agency headquartered in

The U.S. Department of Homeland Security is a federal agency headquartered in

The Securities and Exchange Commission is a federal agency headquartered in

The National Aeronautics and Space Administration is a federal agency

The U.S. Department of Energy is a federal agency headquartered in Washington,

D.C.

34. The Federal Reserve Board of Governors is a federal agency headquartered in

Washington, D.C.

35. The National Endowment for the Arts is a federal agency headquartered in

Washington, D.C.

36. The Executive Office of the President is a federal agency headquartered in

Washington, D.C.

37. Peter B. Hegseth is the Secretary of War and head of Defendant U.S. Department

of War. He is sued in his official capacity.

38. Scott Bessent is the Secretary of the Treasury and head of Defendant U.S.

Department of the Treasury. He is sued in his official capacity.

39. William J. Pulte is the Director of U.S. Federal Housing and head of Defendant

Federal Housing Finance Agency. He is sued in his official capacity.

40. Marco Rubio is the Secretary of State and head of Defendant U.S. Department of

State. He is sued in his official capacity.

- 7 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 11 of 48

41. Robert F. Kennedy, Jr. is the Secretary of Health and Human Services and head of

Defendant U.S. Department of Health and Human Services. He is sued in his official capacity.

42. Howard Lutnick is the Secretary of Commerce and head of Defendant U.S.

Department of Commerce. He is sued in his official capacity.

43. Douglas A. Collins is the Secretary of Veterans Affairs and head of Defendant

U.S. Department of Veterans Affairs. He is sued in his official capacity.

44. Edward C. Forst is the Administrator of Defendant General Services

Administration. He is sued in his official capacity.

45. Scott Kupor is the Director of Defendant U.S. Office of Personnel Management.

He is sued in his official capacity.

46. Ho K. Nieh is the Chairman of Defendant U.S. Nuclear Regulatory Commission.

He is sued in his official capacity.

47. Frank J. Bisigano is the Commissioner of Defendant U.S. Social Security

Administration. He is sued in his official capacity.

48. Kristi Noem is the Secretary of Homeland Security and the head of Defendant

U.S. Department of Homeland Security. She is sued in her official capacity.

49. Paul S. Atkins is the Chairman of Defendant Securities and Exchange

Commission. He is sued in his official capacity.

50. Jared Isaacman is the Administrator of Defendant National Aeronautics and Space

Administration. He is sued in his official capacity.

51. Chris Wright is the Secretary of Energy and head of Defendant U.S. Department

of Energy. He is sued in his official capacity.

52. Jerome H. Powell is the Chairman of Defendant Federal Reserve Board of

Governors. He is sued in his official capacity.

53. Mary Anne Carter is the Chairman of Defendant National Endowment for the

Arts. She is sued in her official capacity.

54. Doe Defendants 1 through 10 are federal departments, agencies, offices, or

instrumentalities—including responsible officials within them—beyond those specifically

- 8 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 12 of 48

identified above that have participated in the development and implementation of the Challenged

Actions. All individual officials among the Doe Defendants are sued in their official capacities.

Their true names and capacities are unknown to Anthropic at this time, and Anthropic will seek

leave to amend this Complaint to identify them as their identities and roles become known.

JURISDICTION AND VENUE

55. This Court has subject-matter jurisdiction under 28 U.S.C. § 1331 because this

civil action arises under the Constitution of the United States and federal statutes. This Court is

authorized to award the requested relief under Rules 57 and 65 of the Federal Rules of Civil

Procedure; the Administrative Procedure Act (APA), 5 U.S.C. §§ 702, 705, 706; the Declaratory

Judgment Act, 28 U.S.C. §§ 2201-02; the All Writs Act, 28 U.S.C. § 1651; and the court’s

inherent equitable powers. The APA waives sovereign immunity. 5 U.S.C. § 702.

56. This Court also has authority to enjoin unlawful official action that is ultra vires,

see, e.g., Reich, 74 F.3d at 1327-28, or that violates the Constitution, see Free Enter. Fund v.

Pub. Co. Acct. Oversight Bd., 561 U.S. 477, 491 n.2 (2010). The Supreme Court has long held

that federal courts have equitable power to grant injunctive relief “with respect to violations of

federal law by federal officials.” Armstrong v. Exceptional Child Ctr., Inc., 575 U.S. 320, 326-27

(2015); see also Larson v. Domestic & Foreign Com. Corp., 337 U.S. 682, 689 (1949).

57. Venue is proper in this District under 28 U.S.C. § 1391(e)(1)(C), because

Defendants are agencies of the United States and officers of the United States acting in their

official capacities, Plaintiff resides in this District, and no real property is involved.

FACTUAL BACKGROUND

Artificial Intelligence (AI) Models

58. Claude is a versatile, industry-leading large language model (LLM) that can be

used in many different contexts depending on a user’s needs. Anthropic first launched Claude in

March 2023. The company has released several more versions of Claude since then, most

recently Claude Opus 4.6 and Claude Sonnet 4.6 in February 2026.

59. LLMs like Claude are algorithmic systems trained on massive datasets to identify

patterns and associations in language, and to generate outputs and take actions that resemble

- 9 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 13 of 48

human responses and actions. Through training, models acquire predictive power and the

transformative ability to take a range of actions in a fraction of the time it would take humans to

perform them.

60. When deployed through a chatbot interface, Claude can interpret and respond to a

vast variety of user inputs, known as “prompts,” in an intelligent, human-like way. Depending on

the nature of the user’s prompt, Claude can: process basic instructions and logical scenarios; take

direction on tone and “personality” when providing outputs; write in different languages; provide

outputs in a variety of programming languages; analyze large amounts of information; and

provide answers to user queries, with detailed background on technical, scientific, and cultural

knowledge.

61. Claude may also be configured with tools that enable it to behave “agentically,”

meaning it can take actions on behalf of a user such as retrieving information, navigating online

resources, writing and executing code, interacting with external services, or carrying out open-

ended tasks that Claude plans and adapts. In certain configurations, Claude can perform tasks

with minimal ongoing user input, operating with a degree of autonomy. Although this agentic

use of AI systems is of particular interest to some users, including governments, it also presents

heightened risks compared to traditional, prompt-response chatbot interactions.

62. AI models like Claude are not perfect. Despite developers’ best efforts, models

can generate inaccurate or misguided responses, or they can “hallucinate,” confidently providing

incorrect information. This is in part because models generate responses by sampling from a

probability distribution rather than by selecting outputs pursuant to predefined rules. As a result,

the outputs may or may not be factually accurate, and the same model, given the same prompt

twice, may provide two different responses.

Anthropic’s Foundational Commitment To AI Safety

63. Anthropic was founded in 2021 by seven former employees of OpenAI

committed to the belief that AI will have a vast impact on the world and that AI development

should maximize positive outcomes for humanity. Anthropic believes that AI policy decisions in

the next few years will touch nearly every part of public life and that questions of AI policy

- 10 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 14 of 48

governance are inherently nonpartisan. To that end, Anthropic has earned a reputation as an

advocate dedicated to building a safer AI ecosystem. In keeping with that founding mission,

Anthropic also builds frontier AI systems and strives to deploy those systems responsibly, in

service of human progress.2 Anthropic began as a research-first company, devoted to AI

research, adversarial testing, and policy work to further AI safety. That focus continues today.

64. As a public benefit corporation (PBC), Anthropic balances stockholder interests

with its public benefit purpose of responsibly developing and maintaining advanced AI for the

long-term benefit of humanity. The Delaware PBC statute permits its board to consider safety,

ethics, and societal impact as part of ordinary corporate decision-making, rather than treat profit

maximization as the sole objective.

65. These beliefs are fully compatible with responsible use of Claude by the

Department of War. Claude has a wide range of specialized defense applications, including

autonomously completing complex software engineering projects related to offensive and

defensive cyber operations and vulnerability detection; supporting military operations;

performing intelligence analysis; and even handling national security workflows on a custom

fine-tuned version of Claude developed for classified networks.

66. Anthropic has developed a detailed Usage Policy to address the unique risks of

AI, encourage safe and responsible uses of its models, and prohibit a wide range of behaviors

contrary to its mission and values. Among other things, that Policy prohibits users from selling

illegal drugs, engaging in human trafficking, exploiting cyber vulnerabilities, designing weapons

or delivery processes for the deployment of weapons, or engaging in surveillance of persons

without their consent. By its terms, the Policy has always prohibited the use of Anthropic’s

services for lethal autonomous warfare without human oversight and surveillance of Americans

en masse.

2 A statement from Dario Amodei on Anthropic’s commitment to American AI leadership,

Anthropic (Oct. 21, 2025), https://www.anthropic.com/news/statement-dario-amodei-american-

ai-leadership.

- 11 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 15 of 48

The Federal Government’s Embrace Of AI And Contracts With Anthropic

67. Since taking office, the Trump Administration has made global adoption of U.S.-

developed AI systems a stated policy priority. The President has issued multiple Executive

Orders focused on America’s global AI dominance.3 His Administration released an “AI Action

Plan” focused in part on promoting AI adoption throughout the federal government, which

Anthropic publicly supported. Last year, the General Services Administration (GSA) added

Claude and other AI providers to its list of approved vendors. The Department likewise has

significantly expanded its use of artificial intelligence and entered into multiple major contracts

with leading AI companies to scale AI capabilities across defense and intelligence missions,

4

including “warfighting, intelligence, business, and enterprise information systems.”5

68. Anthropic is committed to these objectives and has invested considerable

resources to support the government’s national security work. Today, Claude is reportedly the

Department’s most widely deployed and used frontier AI model—and the only one currently on

classified systems.6

69. This did not happen overnight. Anthropic began building the infrastructure,

partnerships, regulatory approvals, and capabilities necessary to support U.S. government

operations in 2023. It joined the AI Safety Institute Consortium, collaborating with the federal

government on AI safety research and evaluation frameworks. It entered into strategic

partnerships with cloud providers to support its growing role in the national security ecosystem.

And it invested substantial resources into pursuing—and obtaining—authorization in the Federal

3 See Exec. Order 14179, 90 Fed. Reg. 8741 (Jan. 23, 2025); Exec. Order 14320, 90 Fed. Reg.

35393 (July 23, 2025).

4 CDAO Announces Partnerships with Frontier AI Companies to Address National Security

Mission Areas, CDAO (July 14, 2025), https://www.ai.mil/latest/news-press/pr-

view/article/4242822/cdao-announces-partnerships-with-frontier-ai-companies-to-address-

national-secu/.

5 Id.

6 See Sheera Frenkel & Julian E. Barnes, Defense Dept. and Anthropic Square Off in Dispute

Over A.I. Safety, N.Y. Times (Feb. 18, 2026),

https://www.nytimes.com/2026/02/18/technology/defense-department-anthropic-ai-safety.html.

- 12 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 16 of 48

Risk and Authorization Management Program (FedRAMP), the government’s security

authorization framework for cloud products and services.

70. Anthropic has also developed specialized “Claude Gov” models tailored

specifically for the national security context.7 These models have been built based on direct

feedback from national security agencies to address real-world requirements, like improved

handling of classified information, enhanced proficiency in critical languages, and sophisticated

analysis of cybersecurity data. Claude Gov models undergo rigorous safety testing consistent

with Anthropic’s commitment to responsible AI.

71. To make Claude more useful for the military and intelligence components of the

federal government, Anthropic does not impose the same restrictions on the military’s use of

Claude as it does on civilian customers. Claude Gov is less prone to refuse requests that would

be prohibited in the civilian context, such as using Claude for handling classified documents,

military operations, or threat analysis. Anthropic’s terms in its existing contracts with the

government also recognize the government’s unique needs and capabilities. For example,

Anthropic’s government-specific addendum to the Usage Policy permits Claude to be used to

analyze lawfully collected foreign intelligence information, which would not be permitted under

the Usage Policy for civilian users.

72. Since 2024, Anthropic has partnered with other national security contractors.

Those partnerships have enabled the incorporation of Claude into the classified systems of the

Department of War and other agencies. And they have allowed for the use of Claude to support

government operations such as rapid processing of complex data, identifying trends, streamlining

document review, and helping government officials make more informed decisions in time-

sensitive situations.8

7 Claude Gov Models for U.S. National Security Customers, Anthropic (June 6, 2026),

https://www.anthropic.com/news/claude-gov-models-for-u-s-national-security-customers.

8 Anthropic and Palantir Partner to Bring Claude AI Models to AWS for U.S. Government

Intelligence and Defense Operations, Business Wire (Nov. 7, 2024),

https://www.businesswire.com/news/home/20241107699415/en/Anthropic-and-Palantir-Partner-

to-Bring-Claude-AI-Models-to-AWS-for-U.S.-Government-Intelligence-and-Defense-

Operations.

- 13 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 17 of 48

73. Last year, Anthropic entered its first direct agreement with the Department’s

Chief Digital and Artificial Intelligence Office (CDAO). Under that agreement, Anthropic

agreed to work with the Department to scope and develop use cases and, eventually, design a

prototype AI service specifically for the Department’s use. CDAO awarded similar agreements

to Google, OpenAI, and xAI, each with a $200 million ceiling value, as part of its “commercial-

first approach to accelerating DoD adoption of AI.”9

74. Anthropic worked diligently under that agreement, scoping out potential ways

that the Department could best be served by Claude and related Anthropic professional services.

During this period, the Department conveyed to Anthropic that Claude was the best solution for

some of the proposals.

75. In the fall of 2025, Anthropic began negotiations for an additional agreement to

provide a version of Claude on the Department’s “GenAI.mil” AI platform. As part of those

discussions, the Department asked Anthropic to excise its Usage Policy and allow the

Department to use Claude for “all lawful uses.” Because of Anthropic’s commitment to U.S.

national security, Anthropic substantially agreed to the proposal—except in two important

respects.

76. First, Anthropic did not develop Claude (or the specialized Claude Gov models)

to deploy lethal autonomous warfare without human oversight. Claude has not been trained or

tested for that use. At least at present, Claude is simply not capable of performing such tasks

responsibly without human oversight.

77. Second, Anthropic is unwilling to agree to Claude’s use for mass surveillance of

Americans. AI tools like Claude enable collection and analysis of information at speeds and

scales not previously contemplated, posing unique risks for civil liberties given the potential for

errors and misuse. These techniques would have been unimaginable when Congress enacted the

existing frameworks regulating how the Executive Branch may conduct surveillance. AI

9 CDAO Announces Partnerships with Frontier AI Companies to Address National Security

Mission Areas, CDAO (July 14, 2025), https://www.ai.mil/latest/news-press/pr-

view/article/4242822/cdao-announces-partnerships-with-frontier-ai-companies-to-address-

national-secu/.

- 14 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 18 of 48

technology is developing far more rapidly than those legal frameworks. And surveillance

conducted using AI poses significantly greater potential to make mistakes—and to amplify the

effect of any mistakes—than traditional techniques.

78. Allowing Claude to be used to enable the Department to surveil U.S. persons at

scale and to field weapons systems that may kill without human oversight would therefore be

inconsistent with Anthropic’s founding purpose and public commitments. These important

restrictions simply reflect what Anthropic knows to be true about Claude’s limitations.

79. The Usage Policy does not provide Anthropic with any special capabilities to

control, oversee, or second-guess the federal government’s operations or the Department’s

military judgments. Nor does providing Claude to the government as a vendor place Anthropic in

a position to intervene in or impede government decision-making. Indeed, while operating under

the terms of the Usage Policy, the Department never previously raised any issues with its use of

Claude or concerns about Anthropic’s potential interference. Anthropic had only ever received

positive feedback about Claude’s capabilities from its government customers.

The Present Dispute

80. Later in 2025, the discussions regarding an additional agreement about deploying

Claude on the “GenAI.mil” platform morphed into a negotiation over the Department’s use of

Claude more broadly. The Department demanded that—across all ongoing and future

deployments of Claude—Anthropic abandon its Usage Policy and instead allow “all lawful use”

of Claude. As part of these new demands, the Department sent partial contract language

incorporating this term to Anthropic.

81. In early January 2026, Secretary Hegseth issued a memorandum directing the

Department to “[u]nleash experimentation with America’s leading AI models Department-wide”

and execute a series of “Pace-Setting Projects” to accelerate AI adoption. To advance that goal,

the memorandum directed the Department’s procurement office to “incorporate standard ‘any

lawful use’ language into any DoW contract” for AI services within 180 days. Three days later,

Secretary Hegseth delivered remarks explaining that the Department was “blowing up . . .

barriers.”

- 15 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 19 of 48

82. Despite disagreement on the two use restrictions, Anthropic has continued to

reiterate its commitment to providing Claude to serve the United States’ national security

interests and to negotiate in good faith with the Department.

83. But the Department chose a different path. In February 2026, a source inside the

Department told reporters that it was “close” to cutting business ties with Anthropic and

designating Anthropic a “supply chain risk,” a designation that—to Anthropic’s knowledge—has

never before been applied to a domestic company.10 The source said: “It will be an enormous

pain in the ass to disentangle, and we are going to make sure they pay a price for forcing our

hand like this.”11

84. Until the Department raised this threat, no government official had ever raised a

concern with Anthropic about potential supply chain vulnerabilities. On the contrary, the

government has consistently provided the security clearances that are necessary for Anthropic’s

personnel to perform classified work. Those clearances remain in place today. Moreover, in 2024

Anthropic became the first frontier AI lab to collaborate with the Department of Energy to

evaluate an AI model in a Top Secret classified environment.

85. Matters came to a head in a meeting between Secretary Hegseth and Dr. Dario

Amodei, Anthropic’s CEO, on February 24, 2026. Secretary Hegseth presented Anthropic with

an ultimatum. He demanded that Anthropic accede to the Department’s demands within four

days or face one of two apparently contradictory punishments: either the Secretary would purport

to invoke the Defense Production Act to force Anthropic to do as he said, or he would cast

Anthropic out of the defense supply chain altogether as a supposed “supply chain risk.”12

Pentagon officials confirmed in the media that the meeting was not intended to drive resolution,

but rather to intimidate Anthropic.13

10 Dave Lawler et al., Exclusive: Pentagon threatens Anthropic punishment, Axios (Feb. 16,

2026), https://www.axios.com/2026/02/16/anthropic-defense-department-relationship-hegseth.

11 Id.

12 Id.; see also Dave Lawler & Maria Curi, Exclusive: Hegseth gives Anthropic until Friday to

back down on AI safeguards, Axios (Feb. 24, 2026),

https://www.axios.com/2026/02/24/anthropic-pentagon-claude-hegseth-dario.

13 Dave Lawler & Maria Curi, Scoop: Hegseth to meet Anthropic CEO as Pentagon threatens

- 16 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 20 of 48

86. After the February 24 meeting, a senior Pentagon official gave Anthropic “until

5:01pm [Eastern] Friday to get on board with the Department of War . . . . If they don’t get on

board, the Secretary of War will ensure the Defense Production Act is invoked on Anthropic,

compelling them to be used by the Pentagon.”14 The same official added, “the Secretary of War

will also label Anthropic a supply chain risk.” Id. In other words, the official suggested that

Anthropic was both necessary to national defense and—at the same time—an unacceptable risk

to national security.

87. On February 26, Dr. Amodei issued a public statement describing Anthropic’s

adherence to its stated policy. He explained that “Anthropic understands that the Department of

War, not private companies, makes military decisions. We have never raised objections to

particular military operations nor attempted to limit use of our technology in an ad hoc manner.”

He again emphasized that the two restrictions giving rise to the dispute address uses that are

“simply outside the bounds of what today’s technology can safely and reliably do,” and that

Anthropic “cannot in good conscience accede to” the Department’s request. He reiterated that

“[o]ur strong preference is to continue to serve the Department and our warfighters—with our

two requested safeguards in place.” And he promised that, “[s]hould the Department choose to

offboard Anthropic, we will work to enable a smooth transition to another provider, avoiding any

disruption to ongoing military planning, operations, or other critical missions.”15

banishment, Axios (Feb. 23, 2026), https://www.axios.com/2026/02/23/hegseth-dario-pentagon-

meeting-antrhopic-claude (quoting a senior official as saying, “This is not a friendly meeting.

This is a sh*t-or-get-off-the-pot meeting.”).

14 Jared Perlo & Gordon Lubold, Anthropic says U.S. military can use its AI systems for missile

defense (Feb. 25, 2026), https://www.nbcnews.com/tech/security/anthropic-pentagon-us-

military-can-use-ai-missile-defense-hegseth-rcna260534.

15 Statement from Dario Amodei on our discussions with the Department of War, Anthropic (Feb.

26, 2026), https://www.anthropic.com/news/statement-department-of-war.

- 17 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEFCase 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 21 of 48

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

The Government Retaliates Against Anthropic

88. The next day—even before the 5:01 p.m. Eastern deadline—President Trump

posted the Presidential Directive, purporting to direct all federal agencies to immediately cease

all use of Anthropic’s technology: 16

89. Secretary Hegseth immediately followed suit by posting a “final” decision on

social media directing his Department to designate Anthropic a “Supply-Chain Risk to National

Security” and decreeing that, “effective immediately,” “no contractor, supplier, or partner that

16 Exhibit 1, Donald J. Trump (@realDonaldTrump), TruthSocial (Feb. 27, 2026, 12:47 PM PT),

https://truthsocial.com/@realDonaldTrump/posts/116144552969293195.

- 18 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 22 of 48

does business with the United States military may conduct any commercial activity with

Anthropic”:17

90. The Secretarial Order left unclarified who is covered as a “partner,” what it means

to do business “with the United States military” versus the Department more broadly, or what

“commercial activity” is prohibited. Regardless of what these other companies must do, the

17 Exhibit 2, Secretary of War Pete Hegseth (@SecWar), X (Feb. 27, 2026, 2:14 PM PT),

https://x.com/SecWar/status/2027507717469049070.

- 19 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 23 of 48

Order also insisted that “Anthropic will continue to provide the Department of War its services

for a period of no more than six months.”18

91. But the Secretary left no doubt about his reasons: “Anthropic’s stance is

fundamentally incompatible with American principles.”19 According to the Secretary, this

“stance” includes “Silicon Valley ideology,” “corporate virtue-signaling,” “defective altruism,”

“arrogance,” and even an attempt to hold “America’s warfighters . . . hostage [to] the ideological

whims of Big Tech.”20 The Secretary thus distorted Anthropic’s clear-eyed, expertise-driven

understanding of its own technology’s current limits into purported ideological extremism.

92. GSA also took immediate steps in “support of President Trump’s directive,”

which it understood to “rejec[t] attempts to politicize work” and to require federal agencies to

contract only with AI companies “who fit the bill.”21 In a news release issued the same day as the

Presidential Directive, GSA announced that it was removing Anthropic from USAi.gov and the

Multiple Award Schedule contracts. A top GSA official separately announced that the agency

had terminated Anthropic’s “OneGov” contract.22

93. Other government agencies soon fell in line, issuing multiple directives to begin

to implement the President and the Secretary’s directives. For example, the Department of State

and the Department of Health and Human Services (HHS) have acted on the President’s directive

through internal communications, according to public reporting.23 Monday morning, the U.S.

Department of the Treasury and the Federal Housing Finance Agency announced they were

terminating all use of Claude.24 Anthropic also received reports that the Chief Information

18 Id.

19 Id.

20 Id.

21 Press Release, Gen. Servs. Admin., GSA Stands with President Trump on National Security

AI Directive (Feb. 27, 2026), https://www.gsa.gov/about-us/newsroom/news-releases/gsa-stands-

with-president-trump-on-national-security-ai-directive-02272026.

22 Josh Gruenbaum (@FASCommissioner), X (Feb. 27, 2026, 6:21 PM ET),

https://x.com/FASCommissioner/status/2027524519703838973.

23 Lindsey Wilkinson & Madison Alder, Anthropic faces fallout across federal agencies from

DOD clash, FedScoop (Feb. 27, 2026), https://fedscoop.com/anthropic-claude-dod-federal-

agency-fallout-trump-hegseth/.

24 Scott Bessent (@SecScottBessent), X (Mar. 2, 2026, 10:57 AM ET),

- 20 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 24 of 48

Officer of a federal civilian agency advised all non-Department of War leadership to stop using

Claude.

94. Private actors also took heed. Anthropic immediately received outreach from

numerous outside partners—from customers, to cloud providers, to investors—expressing

confusion about what was required of them and concern about their ability to continue to work

with Anthropic. Since the Challenged Actions, dozens of companies have contacted Anthropic

seeking clarity, guidance, and, in some cases, an understanding of their termination rights.

95. An official confirmed that the Department’s actions are a response to Anthropic’s

purported “behavior” in negotiations and threatened not just to terminate Anthropic’s contracts

but “require that all our vendors and contractors certify that they don’t use any Anthropic

models.”25

96. Other government officials relayed the personal and ideological nature of the

Department’s objective: “The problem with [Anthropic’s CEO] Dario [Amodei] is, with him, it’s

ideological. We know who we’re dealing with.”26 This followed public condemnation of

Anthropic and its usage policies by the Department’s Chief Technology Officer as “not

democratic.”27

97. Throughout, the federal government has never once expressed concerns about

Anthropic’s security or Claude’s competencies. Instead, it has repeatedly recognized that

Anthropic is not only safe but an important national asset. Claude’s FedRAMP authorization

represents the highest level of cloud security certification for the handling of unclassified and

https://x.com/secscottbessent/status/2028499953283117283; William Pulte (@pulte), X (Mar. 2,

2026, 11:12 AM ET), https://x.com/pulte/status/2028503809299779866.

25 Morgan Phillips, Maduro raid questions trigger Pentagon review of top AI firm as potential

‘supply chain risk,

’ Fox News (Feb. 16, 2026), https://www.foxnews.com/politics/maduro-raid-

questions-trigger-pentagon-review-top-ai-firm-potential-supply-chain-risk.

26 Dave Lawler & Maria Curi, Scoop: Hegseth to meet Anthropic CEO as Pentagon threatens

banishment, Axios (Feb. 23, 2026), https://www.axios.com/2026/02/23/hegseth-dario-pentagon-

meeting-antrhopic-claude.

27 See Sydney J. Freedberg Jr., Pentagon CTO says it’s ‘not democratic’ for Anthropic to limit

military use of Claude AI, Breaking Defense (Feb. 19, 2026),

https://breakingdefense.com/2026/02/pentagon-cto-says-its-not-democratic-for-anthropic-to-

limit-military-use-of-claude-ai/.

- 21 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 25 of 48

controlled unclassified information. The Department approved (and has continued to maintain) a

facility clearance for Anthropic as well as numerous security clearances for Anthropic’s

personnel so they can perform classified work. Never during any of these security-focused

processes did the government determine that Anthropic or its services posed a supply chain risk.

Indeed, the FedRAMP authorization and facility security clearance and personnel clearances

could not have been issued had any such determination been made.

98. Even during the recent negotiations, the government has repeatedly and publicly

praised Claude’s capabilities. Chief Technology Officer and Under Secretary of War Emil

Michael, while describing the dispute with Anthropic, explicitly characterized Anthropic as one

of America’s “national champions” in AI.28 In the February 24 meeting with Dr. Amodei,

Secretary Hegseth described Anthropic’s technology as having “exquisite capabilities” and stated

that the Department would “love” to work with Anthropic.

99. Senior administration officials have likewise repeatedly acknowledged that

displacing Anthropic from its role would be disruptive because competing AI models “are just

behind” when it comes to specialized government applications.29

100. Department officials have even expressed concerns about the consequences of

losing access to Claude.30 Describing the dispute between Anthropic and the Department, one

official stated that “[t]he only reason we’re still talking to these people is we need them and we

need them now. The problem for these guys is they are that good.” 31

101. Indeed, the President and Secretary Hegseth insisted that Claude must remain

available to the Department for six months—even after another AI company had indicated it

28 Sydney J. Freedberg Jr., Pentagon CTO says it’s ‘not democratic’ for Anthropic to limit

military use of Claude AI, Breaking Defense (Feb. 19, 2026),

https://breakingdefense.com/2026/02/pentagon-cto-says-its-not-democratic-for-anthropic-to-

limit-military-use-of-claude-ai/.

29 Dave Lawler & Maria Curi, Scoop: Hegseth to meet Anthropic CEO as Pentagon threatens

banishment, Axios (Feb. 23, 2026), https://www.axios.com/2026/02/23/hegseth-dario-pentagon-

meeting-antrhopic-claude.

30 See Dave Lawler & Maria Curi, Exclusive: Hegseth gives Anthropic until Friday to back down

on AI safeguards, Axios (Feb. 24, 2026), https://www.axios.com/2026/02/24/anthropic-

pentagon-claude-hegseth-dario.

31 Id.

- 22 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 26 of 48

would accede to the Department’s demand to make its models available for “all lawful uses,” and

apparently as the Department was in talks with a third AI company that recently announced it is

inclined to do the same thing.32 Within hours of the Challenged Actions, moreover, the

Department reportedly “launched a major air attack in Iran with the help of [the] very same

tools” that are “made by” Anthropic and are the subject of the Challenged Actions.33

102. And senior officials within the Department recently confirmed to the press what is

apparent from the facts: One official who manages information security said that the Secretarial

Order was “ideological” rather than an accurate description of risk. Another official, who

specifically evaluates supply chain risk and other potential intelligence threats, acknowledged

“there is no evidence of supply-chain risk” from Anthropic’s AI model and reiterated that the

Secretarial Order was “ideologically driven.”34

103. Indeed, the President himself made clear that his Administration’s retaliatory

actions towards Anthropic were a direct result of the views Anthropic expressed to the

government and the public about the limitations on the use of its own product: “Well, I fired

Anthropic. Anthropic is in trouble because I fired [them] like dogs, because they shouldn’t have

done that.”35

The Secretary Notifies Anthropic Of His “Supply Chain Risk” Designation

104. Even as agencies across the federal government moved to implement the

Presidential Directive, Dr. Amodei and Under Secretary of War Michael continued negotiations

32 Keach Hagey, Altman Says OpenAI Is Working on Pentagon Deal Amid Anthropic Standoff,

Wall St. J. (Feb. 27, 2026), https://www.wsj.com/tech/ai/openais-sam-altman-calls-for-de-

escalation-in-anthropic-showdown-with-hegseth-03ecbac8?mod=Searchresults&pos=1&page=1.

33 Marcus Weisberger, A. Ramkumar, & S. Holliday, U.S. Strikes in Middle East Use Anthropic,

Hours After Trump Ban, Wall St. J. (Feb. 28, 2026), https://www.wsj.com/livecoverage/iran-

strikes-2026/card/u-s-strikes-in-middle-east-use-anthropic-hours-after-trump-ban-

ozNO0iClZpfpL7K7ElJ2?mod=Searchresults&pos=2&page=1.

34 Patrick Tucker, Pentagon’s war on Anthropic based on ‘dubious’ legal thinking and ideology–

not real risk, sources say, Defense One (Mar. 3, 2026),

https://www.defenseone.com/business/2026/03/pentagons-war-anthropic-based-dubious-legal-

thinking-and-ideologynot-real-risk-sources-say/411849/.

35 Dasha Burns, Trump says he’ll help pick Iran’s leader, predicts regime change in Cuba,

Politico (Mar. 5, 2026), https://www.politico.com/news/2026/03/05/trump-unleashed-president-

bullish-on-iran-eyeing-regime-change-in-cuba-and-impatient-with-ukraine-00814292.

- 23 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 27 of 48

in an effort to resolve or de-escalate the dispute. Those discussions were still underway when, at

8:48 p.m. Eastern on March 4, the Secretary of War sent Anthropic a letter. The letter, dated

March 3, 2026, notified it of the “supply-chain risk” designation—almost a week after the

Secretary had announced that designation on social media.

105. The two-page letter did not explain what risk Anthropic’s services supposedly

pose to national security. Its stated rationale reads in full: “the Department of War has

determined that (i) the use of the Covered Entity’s products or services in DoW covered systems

presents a supply chain risk and that the use of the Section 3252 authority to carry out covered

procurement actions is necessary to protect national security by reducing supply chain risk, and

(ii) less intrusive measures are not reasonably available to reduce such supply chain risk.”

106. Based on that “determination,

” the Secretarial Letter purports to exclude

Anthropic—including all of its subsidiaries, successors, and affiliates—as a source for all

Department procurements involving covered national security systems, effective immediately.

36

The Letter does not explain the scope of procurements covered by the Secretary’s action.

The Challenged Actions Are Causing Immediate And Irreparable Harm To Anthropic

107. The Challenged Actions have inflicted immediate, far-ranging, and irreversible

harm on Anthropic. These harms will continue unless the Challenged Actions are declared

unlawful and enjoined.

108. Anthropic has built a reputation as a public benefit corporation that is committed

to AI safety and the responsible deployment of its technology. That reputation is critical to its

continued success and growth. Secretary Hegseth’s unlawful designation of Anthropic as “a

Supply-Chain Risk to National Security” undoubtedly harms Anthropic’s reputation, as does

Defendants’ unlawful decision to bar “EVERY Federal Agency in the United States

Government” from using Anthropic’s technology.

36 At the same time it received the Secretarial Letter, Anthropic also received a letter from the

Secretary (also dated March 3) invoking the supply chain risk designation authority set out in 41

U.S.C. § 4713. By statute, judicial review of actions taken under Section 4713 lies exclusively in

the D.C. Circuit. See 41 U.S.C. § 1327(b). Anthropic intends to challenge that separate action in

that forum and does not raise claims related to it in this action.

- 24 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 28 of 48

109. The Challenged Actions also inflict immediate and unrecoverable revenue losses:

Anthropic stands to lose the federal contracts it already has, as well as its prospects to pursue

federal contracts in the future.

110. Anthropic’s business partnerships and contracts with other federal contractors are

likewise in jeopardy. For example, one federal contractor with whom Anthropic has built custom

applications has indicated that it may suspend that work or even remove Claude from existing

deployments. Other federal contractors are raising concerns, pausing collaborations, and

considering terminating contracts. Anthropic has no way to obtain redress from the government

for those economic harms.

111. And those practical and economic injuries are not the only irreparable harms

inflicted by the Challenged Actions. “The loss of First Amendment freedoms, for even minimal

periods of time, unquestionably constitutes irreparable injury.” Roman Catholic Diocese of

Brooklyn v. Cuomo, 592 U.S. 14, 19 (2020) (per curiam).

112. All of this is precisely what Defendants intended: to punish Anthropic for

adhering to its views. Anthropic was founded on its commitment to developing AI responsibly.

Defendants presented Anthropic with a stark choice: silence its views on safe AI, capitulate to

the Department’s demands, and offer Claude on terms that are unsafe and violate its core

principles—or else suffer swift harm at the hand of the federal government. When Anthropic

adhered to its longstanding views about AI safety and the limitations of its services, Defendants

carried out that threat.

CLAIMS

COUNT I

ADMINISTRATIVE PROCEDURE ACT; 10 U.S.C. § 3252

(5 U.S.C. § 706)

(DEFENDANTS HEGSETH AND THE DEPARTMENT OF WAR)

113. 114. Anthropic incorporates by reference the allegations of the preceding paragraphs.

The APA requires courts to “hold unlawful and set aside” final agency action that

is “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law,” or is

- 25 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 29 of 48

“in excess of statutory jurisdiction, authority, or limitations, or short of statutory right,” or

“without observance of procedure required by law.” 5 U.S.C. § 706(2)(A), (C), (D).

115. The February 27 Secretarial Order purported to “direct[] the Department of War

to designate Anthropic a Supply-Chain Risk to National Security” and ordered that, “[e]ffective

immediately, no contractor, supplier, or partner that does business with the United States military

may conduct any commercial activity with Anthropic.” The Order also emphasized that “[t]his

decision is final.”37

116. The Secretarial Order is a final agency action for purposes of the APA. It is an

“agency action” because it is an “order” (i.e., a “disposition . . . in a matter other than

rulemaking”) and also a “sanction” that “prohibit[s],” “limit[s],” or otherwise “affect[s]”

Anthropic’s freedom to compete for federal contracts and maintain its business relationships.

5 U.S.C. § 551(6), (10), (13). It is final both because Secretary Hegseth said so and because it

finally “determine[s]” the “rights or obligations” of Anthropic and is backed by “legal

consequences.” Bennett v. Spear, 520 U.S. 154, 177-78 (1997). Effective “immediately,” the

decision purports to direct that no contractor, supplier, or partner that does business with the

United States military may conduct any commercial activity with Anthropic.

117. A week later, the Secretary sent Anthropic a letter notifying it that the Department

“has determined” that the use of Anthropic’s “products or services in DoW covered systems

presents a supply chain risk” and that it is necessary for the Department to use its authority under

10 U.S.C. § 3252 “to protect national security by reducing supply chain risk.” The Secretarial

Letter also asserts that “less intrusive measures are not reasonably available to reduce such

supply chain risk.” Those statements are the only explanations offered in the Secretarial Letter

for the supply chain risk designation. And the Secretarial Letter does not purport to rescind or

amend the Secretarial Order. See generally Nat’l Urb. League v. Ross, 508 F. Supp. 3d 663 (N.D.

Cal. 2020) (“A final agency action does not become non-final after it is implemented.”).

37 Exhibit 2, Secretary of War Pete Hegseth (@SecWar), X (Feb. 27, 2026, 2:14 PM PT),

https://x.com/SecWar/status/2027507717469049070.

- 26 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 30 of 48

118. An agency acts arbitrarily and capriciously when it “entirely fail[s] to consider an

important aspect of the problem,” offers “an explanation for its decision that runs counter to the

evidence before the agency,” or fails to “articulate a satisfactory explanation for its action

including a rational connection between the facts found and the choice made.” Motor Vehicle

Mfrs. Ass’n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983) (internal quotation marks

omitted).

119. The Secretarial Order, and the attempt to implement and explain that Order via

the Secretarial Letter, violates the standards of Section 706 at every turn.

120. First, the Order exceeds the authority granted by Congress in 10 U.S.C. § 3252,

the federal statute addressing “supply chain risk[s].” That statute does not provide the

government a remedy for failed contract negotiations. Nor does it delegate freewheeling

authority to the Secretary to redefine “supply chain risk” to cover a contractor who declines to

modify its terms of use to track the Department’s preferences.

121. Instead, Section 3252 authorizes exclusion with respect to a prime contractor or

subcontractor only when necessary to protect against the risk that an adversary may “sabotage

. . . or otherwise subvert” an information system used for national security purposes. 10 U.S.C.

§ 3252(b)(2)(A), (d)(4)-(5); 44 U.S.C. § 3552(b)(6). The Secretary has not determined, and

cannot reasonably determine, that Anthropic’s services present a risk of sabotage or subversion

by an adversary to the United States.

122. Anthropic is not, and has no ties to, an “adversary” to the United States. The

Executive Branch has defined the term to mean China, Russia, Iran, North Korea, Cuba, and

Venezuela. See Exec. Order No. 13,873, 84 Fed. Reg. 22689 (May 15, 2019); 15 C.F.R.

§ 791.4(a). Anthropic is a U.S.-incorporated, U.S.-headquartered public benefit corporation with

a demonstrated history of supporting the United States government and its national security

interests. The Secretary has not articulated any determination otherwise. Nor is there any other

valid basis for the Secretary to determine that designating Anthropic presents a risk of

“sabotage” or “subver[sion].” Indeed, Anthropic has gone to significant lengths to prevent the

use of its technology by entities linked to the Chinese Communist Party, has shut down attempts

- 27 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 31 of 48

to abuse Claude for state‑sponsored cyber operations, and has advocated for strong export

controls on the most powerful chips used to train AI, all to preserve the U.S. lead in frontier AI

development.

123. Second, the Secretary’s actions failed to follow the procedure Congress required

before excluding from contracts or subcontracts on the basis that it poses an unacceptable

“supply chain risk.” Under Section 3252, the Secretary must consult with other relevant officials

and determine in writing (1) that an exclusion is “necessary to protect national security by

reducing supply chain risk,” and (2) that “less intrusive measures are not reasonably available to

reduce such supply chain risk.” 10 U.S.C. § 3252(b)(1), (b)(2)(A)-(B). Then the Secretary must

notify the appropriate congressional committees of that determination, providing a summary of

the risk assessment and the basis for determining that less intrusive options were not available.

10 U.S.C. § 3252(b)(3). On information and belief, no valid Section 3252 determination was

made prior to the February 27 Secretarial Order. The Secretary did not consult with relevant

procurement officials, did not make any written determination that less intrusive measures were

unavailable, and did not notify Congress before issuing the Order. And even the Secretarial

Letter received by Anthropic on March 4, which recited the “necessary to protect national

security” and “less intrusive measures are not reasonably available” language from 10 U.S.C.

§ 3252(b)(2)(A)-(B), did not describe any consultation with relevant procurement officials or any

congressional notification.

124. With respect to contracts entered directly with the government, Section 3252

authorizes the exclusion of a source only if it has failed either to “meet qualification standards”

or “achieve an acceptable rating with regard to an evaluation factor.” 10 U.S.C. § 3252(d)(2)(A)-

(B). In both cases, those conditions relate to the risk that an adversary may sabotage, maliciously

interfere with, or otherwise subvert a covered system. The Secretary has not determined—and

could not reasonably determine—that Anthropic’s services fail to meet qualification standards or

achieve an acceptable rating related to any evaluation factor for a procurement. The February 27

Secretarial Order contains no such determination. And the Secretarial Letter sent on March 4

does not address those statutory criteria.

- 28 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 32 of 48

125. To the contrary, the Secretary himself has recognized Claude’s capabilities as

“exquisite.” His Department suggested that Claude was so vital to our national defense that it

needed to be commandeered under the Defense Production Act. And he has ordered that

“Anthropic will continue to provide” its services to the Department of War for up to “six

months.” The “unexplained inconsistenc[y]” between simultaneously designating Anthropic’s

services a supply chain risk vulnerable to “sabotage” or other “subver[sion]” by a foreign

adversary while directing those services to be used for up to six months for national security

purposes demonstrates the arbitrariness of the Secretary’s final decision. Dist. Hosp. Partners,

L.P. v. Burwell, 786 F.3d 46, 59 (D.C. Cir. 2015) (collecting authority).

126. Additionally, nothing in the statute authorizes the Secretary to require every

“contractor, supplier, or partner that does business with the United States military” to blacklist

the excluded source.

127. Third, the Secretarial Order was arbitrary and capricious because it failed to

provide a rational and “satisfactory explanation” for designating Anthropic a supply chain risk.

Motor Vehicle Mfrs. Ass’n, 463 U.S. at 43. The Secretary’s February 27 Order announcing his

“final” decision contains invective against Anthropic, but no explanation of why Claude

constitutes a supply chain risk. It does not attempt to reconcile the Secretary’s assertion that

those models are a threat “to National Security” with his decision to allow the Department to

continue using them for half a year—let alone the Department’s past praise for those models or

its simultaneous suggestion that Anthropic might be commandeered into providing them on the

Department’s terms under the Defense Production Act.

128. The post hoc Secretarial Letter does not meaningfully elaborate on that

explanation. It parrots the statutory predicates of Section 3252: that Anthropic presents a “supply

chain risk,” that the designation is “necessary to protect national security,” and that “less

intrusive measures [were] not reasonably available.

” But it offers no explanation for any of these

conclusions; addresses none of the inconsistencies that rendered the Secretarial Order arbitrary;

and supplies none of the reasoned analysis the Order lacked.

- 29 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 33 of 48

129. The only explanation provided by the Secretary for his action is pure retaliation.

That is plain on the face of the Secretarial Order, in which the Secretary criticized Anthropic as

“ideological” and insufficiently “patriotic.” And it is confirmed by senior Department officials

who unabashedly told the press that the Secretary designated Anthropic as a supply chain risk to

“make sure [Anthropic] pays a price” for declining to concede to the Department’s demands;38

that the Secretarial Order was “ideological” rather than an accurate description of risk; that

“there is no evidence of supply-chain risk”; and that the Secretarial Order was “ideologically

driven.”39

130. The Secretary’s actions are arbitrary and capricious in multiple other ways. For

example, the Secretary failed to consider less restrictive alternatives. Several were available here,

and they had been offered as options by Anthropic itself. First, Anthropic repeatedly offered the

Department that it would support an orderly transition to a new provider—one willing to accept

the Department’s proposed terms—at nominal cost if the parties failed to come to an agreement.

But the Department had other options as well, including agreeing to Anthropic’s proposed usage

limitations; or continuing the negotiations already underway. Neither the Secretarial Order nor

the Secretarial Letter identifies any of these alternatives, much less explains why they are

insufficient.

131. The Secretary also failed to address the consequences of its actions for Anthropic,

other companies that deal with the federal government, and Anthropic’s commercial

counterparties. He also failed to reasonably account for Anthropic’s reliance interests. Neither

the Secretarial Order nor the Secretarial Letter grapples with those considerations. And the

Secretarial Order relied on extra-statutory factors that Congress did not intend for him to

38 Dave Lawler et al., Exclusive: Pentagon threatens Anthropic punishment, Axios (Feb. 16,

2026), https://www.axios.com/2026/02/16/anthropic-defense-department-relationship-hegseth.

39 Patrick Tucker, Pentagon’s war on Anthropic based on ‘dubious’ legal thinking and ideology–

not real risk, sources say, Defense One (Mar. 3, 2026),

https://www.defenseone.com/business/2026/03/pentagons-war-anthropic-based-dubious-legal-

thinking-and-ideologynot-real-risk-sources-say/411849/.

- 30 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 34 of 48

consider under Section 3252, such as Anthropic’s position in contract negotiations and its public

statements on AI safety.

132. For these reasons, the Court should declare that the Secretarial Order is “in excess

of statutory jurisdiction, authority, or limitations,” 5 U.S.C. § 706(2)(C), and “arbitrary,

capricious . . . or otherwise not in accordance with law,” id. § 706(2)(A), set the order aside, and

enjoin Defendants (other than the President) from taking any action to implement or enforce it,

including through the Secretarial Letter.

133. Defendants’ APA violations have caused Anthropic ongoing and irreparable

harm.

COUNT II

FIRST AMENDMENT TO THE U.S. CONSTITUTION

(EQUITABLE CAUSE OF ACTION; 5 U.S.C. § 702)

(ALL DEFENDANTS)

134. Anthropic incorporates by reference the allegations of the preceding paragraphs.

135. The First Amendment to the Constitution provides that the federal Government

“shall make no law . . . abridging the freedom of speech . . . or [abridging] the right of the people

to petition the Government for a redress of grievances.” U.S. Const. amend. I.

136. The Challenged Actions violate Anthropic’s First Amendment rights because they

constitute paradigmatic retaliation against Anthropic’s expressive activities, including protected

speech, protected viewpoints, and protected petitioning of the government.

137. The First Amendment “prohibits government officials from subjecting individuals

to retaliatory actions after the fact for having engaged in protected speech.” Hous. Cmty. Coll.

Sys. v. Wilson, 595 U.S. 468, 474 n.2 (2022); Nieves v. Bartlett, 587 U.S. 391, 398 (2019)

(similar). Indeed, “[s]tate action designed to retaliate against and chill” protected expression

“strikes at the heart of the First Amendment.” Gibson v. United States, 781 F.2d 1334, 1338 (9th

Cir. 1986).

138. Succeeding on a retaliation claim requires Anthropic to show that “(1) [it] was

engaged in a constitutionally protected activity, (2) the defendant’s actions would chill a person

- 31 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 35 of 48

of ordinary firmness from continuing to engage in the protected activity and (3) the protected

activity was a substantial or motivating factor in the defendant’s conduct.” O’Brien v. Welty, 818

F.3d 920, 932 (9th Cir. 2016); President & Fellows of Harvard Coll. v. United States Dep’t of

Homeland Sec., 788 F. Supp. 3d 182, 206 (D. Mass. 2025) (“The elements of a Petition Clause

retaliation claim are identical to those of a free speech retaliation claim.”). All three elements are

easily established here.

139. First, Anthropic engaged in protected First Amendment expression, in multiple

respects.

140. To start, Anthropic has been a leading voice on AI safety and policy since its

inception. The company frequently weighs in on pending legislation: It has advocated for the

bipartisan Future of AI Innovation Act, which supports the efforts of the National Institute of

Standards and Technology’s Center for AI Standards and Innovation (CAISI) to undertake

research on AI safety risks. And it has backed the CREATE AI Act of 2025 and the GAIN Act of

2025—bipartisan safety bills that align with the company’s policy priorities. Anthropic also

maintains a bipartisan lobbying effort and has donated money to organizations that promote AI

safety.

141. The company’s public speech extends to its Usage Policy. That policy, posted

prominently on Anthropic’s website, implements and embodies the company’s foundational

commitment to the safe and responsible use of AI. Consistent with Anthropic’s founding ethos,

the policy “is calibrated to strike an optimal balance between enabling beneficial uses and

mitigating potential harms.”40 As explained above, the Usage Policy has never permitted Claude

to be used for mass surveillance of Americans or for lethal autonomous warfare.

142. Anthropic’s executives speak publicly on these topics. In June 2025, Dr. Amodei

published an op-ed opposing federal legislation that would have imposed a moratorium on state

regulation of AI.41 In October 2025, he released a statement praising President Trump’s AI

40 Usage Policy, Anthropic, https://www.anthropic.com/legal/aup (last visited Mar. 7, 2026).

41 Dario Amodei, Anthropic C.E.O.: Don’t Let A.I. Companies off the Hook, N.Y. Times:

Opinion (June 5, 2025), https://www.nytimes.com/2025/06/05/opinion/anthropic-ceo-regulate-

transparency.html.

- 32 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 36 of 48

action plan, reiterating his opposition to a federal moratorium on state AI regulation, and

emphasizing Anthropic’s support for SB 53, a since-enacted California AI safety bill.42 And, as

noted above, on February 26, 2026, he issued a public statement regarding the importance of

Anthropic’s usage restrictions on lethal autonomous warfare and mass surveillance of

Americans, emphasizing that those uses are “simply outside the bounds of what today’s

technology can safely and reliably do,” and that Anthropic “cannot in good conscience” abandon

those particular restrictions.

143. In addition, Anthropic’s communications with the government are protected

speech. Cf. Janus v. Am. Fed’n of State, Cnty., & Mun. Emps., Council 31, 585 U.S. 878, 893-94

(2018) (recognizing that “collective bargaining” with the government is “private speech” that is

protected by the First Amendment); President & Fellows of Harvard Coll., 788 F. Supp. 3d at

203 (“refusing to cede” on issues of public importance “constitute[s] . . . protected conduct” even

if expressed as “rejection” of contract terms).

144. Throughout its negotiations with the Department, Anthropic expressed its views

about Claude’s capabilities and the uses to which Claude can safely and responsibly be put.

Anthropic has also spoken out about the threat to civil liberties that AI-enabled mass surveillance

of Americans poses. Anthropic has discussed these issues directly with the Department and has

shared its views with the public.43 These expressions of Anthropic’s viewpoints are entitled to

full First Amendment protection. And that expression is what the Challenged Actions seek to

punish.

145. Anthropic also engaged in protected First Amendment activity when it petitioned

the government to honor Anthropic’s use restrictions with respect to lethal autonomous warfare

systems that lack any human oversight and mass surveillance of Americans. The First

Amendment protects the right “to petition the Government for a redress.” U.S. Const. amend. I.

42 A statement from Dario Amodei on Anthropic’s commitment to American AI leadership,

Anthropic (Oct. 21, 2025), https://www.anthropic.com/news/statement-dario-amodei-american-

ai-leadership.

43See, e.g., Statement from Dario Amodei on our discussions with the Department of War,

Anthropic (Feb. 26, 2026), https://www.anthropic.com/news/statement-department-of-war.

- 33 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 37 of 48

Anthropic exercised that right by communicating its position to the Department, explaining the

basis for that position, and seeking to persuade the government to embrace that view. See BE &

K Const. Co. v. N.L.R.B., 536 U.S. 516, 525 (2002) (“[T]he right to petition extends to all

departments of the Government”) (citation omitted)). Anthropic was not simply engaged in

contract negotiations; it was expressing a position on an issue of significant public importance

for which it had unique expertise—the appropriate use of its own AI models. The government’s

response was drastic and punitive, retaliating against the core freedoms the Petition Clause

protects.

146. Second, the Challenged Actions impose significant financial and reputational

costs on Anthropic that would chill a company of ordinary firmness from continuing to engage in

expressive activity. Government action is “adverse” for purposes of a First Amendment

retaliation claim if it is “designed to . . . chill political expression,” Mendocino Env’t Ctr. v.

Mendocino Cnty., 14 F.3d 457, 464 (9th Cir. 1994) (emphasis added), or “would chill a person of

ordinary firmness from continuing to engage in the protected activity,” Blair v. Bethel Sch. Dist.,

608 F.3d 540, 543 (9th Cir. 2010). The Challenged Actions satisfy both tests. By their very

terms, they are intended to force Anthropic to “get their act together[] and be helpful.”44 And

they carry severe and wide-ranging consequences that ripple far beyond any single contract.

147. The Challenged Actions also assign Anthropic a “supply chain risk” designation

that is reserved for companies that create a risk of “sabotage” or other acts of subversion by a

foreign “adversary.” 10 U.S.C. § 3252(d)(4). That label will follow Anthropic into every future

procurement relationship across the federal government and with federal contractors, not to

mention relationships with states and local governments and customers in other sectors. The

threat of that extraordinarily stigmatizing label would undoubtedly chill the expressive activities

of a company of ordinary firmness.

148. This adversity is severe, particularly in the fiercely competitive AI marketplace,

where reputational damage can quickly lead to pecuniary harm. See Riley’s Am. Heritage Farms

44 Exhibit 1, Donald J. Trump (@realDonaldTrump), TruthSocial (Feb. 27, 2026, 12:47 PM PT),

https://truthsocial.com/@realDonaldTrump/posts/116144552969293195.

- 34 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 38 of 48

v. Elsasser, 32 F.4th 707, 723 (9th Cir. 2022) (“A plaintiff establishes . . . adverse action . . . by

demonstrating that the government action threatened or caused pecuniary harm”); Arizona

Students’ Ass’n v. Arizona Bd. of Regents, 824 F.3d 858, 868 (9th Cir. 2016) (“[T]he government

may chill speech by threatening or causing pecuniary harm . . . [or] withholding a license, right,

or benefit . . . .”).

149. Third, Anthropic’s protected expression was not only a substantial factor

underlying the Challenged Actions, it was the motivating factor. The causal link could not be

clearer: Defendants threatened Anthropic and then took the Challenged Actions only after

Anthropic refused to change its position on acceptable uses of Claude and publicly explained

why. Indeed, the government made clear that it took the Challenged Actions because of

Anthropic’s steadfast expression of its views about what Claude can and cannot do. For example,

Secretary Hegseth directly criticized Anthropic’s “rhetoric” when he announced the supply chain

action and faulted the company for not being “more patriotic.”

150. Actions designed to punish ideological disagreement are necessarily motivated by

protected First Amendment activity. See, e.g., Mendocino Envtl. Ctr., 14 F.3d at 464; see also

Perkins Coie LLP v. U.S. Dep't of Just., 783 F. Supp. 3d 105 (D.D.C. 2025) (holding Executive

Order 14230 unconstitutional as a retaliation for protected speech because its text made “clear

that President Trump and his administration disfavor the specific messages conveyed by

plaintiff”).

151. And Defendants’ public statements confirm that the government took the

Challenged Actions because of what Anthropic said, not because of any legitimate procurement

or security concern. No government actor has ever even attempted to identify any technical

deficiency in Claude. To the contrary, Claude has instead been an unmitigated success for the

American military. Perhaps that is why the government initially threatened to invoke the Defense

Production Act against Anthropic and compel it to provide the very service that the government

now calls a supply chain risk. In the government’s own words, “we need them and we need them

- 35 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 39 of 48

now” because Claude is just “that good.”45 Without any technical motivations supporting the

Challenged Actions, the only motivation left is the one candidly expressed by Defendants:

disagreement with Anthropic’s views.

152. To be sure, if it complies with the Constitution and governing statutes and

regulations, the Department may terminate its contract with Anthropic. And it may look to

procure services from other AI companies on the terms it prefers, as it has already done.

Exercising that authority would have been unremarkable. Anthropic even offered to facilitate

such a transition. But the Challenged Actions took a different path. These needless and

extraordinarily punitive actions, imposed in broad daylight, are a paradigm of unconstitutional

retaliation. See Soranno’s Gasco’s Inc. v. Morgan, 874 F.2d 1310, 1316 (9th Cir. 1989)

(inferring a retaliatory motivation where the government’s “chosen course of action was

designed to maximize harm”).

153. The government’s First Amendment retaliation is made worse by the fact that it is

content- and viewpoint-based. It is content-based because the retaliation is targeted at Anthropic

for speaking on issues of AI safety and responsible AI use—“speech on public issues” that

“occupies the highest rung of the hierarchy of First Amendment values.” Snyder v. Phelps, 562

U.S. 443, 452 (2011). The Challenged Actions also punish Anthropic not just for speaking on

that topic, but for Anthropic’s viewpoints on that topic. See, e.g., Pleasant Grove City v.

Summum, 555 U.S. 460, 469 (2009) (“restrictions based on viewpoint are prohibited”).

154. Defendants’ content- and viewpoint-based acts are subject to, but cannot possibly

satisfy, strict scrutiny. See e.g., Vidal v. Elster, 602 U.S. 286, 293 (2024); Waln v. Dysart Sch.

Dist., 54 F.4th 1152, 1162 (9th Cir. 2022) (“Viewpoint-based restrictions on speech are subject

to strict scrutiny.”).

155. To survive strict scrutiny, the government must adopt “the least restrictive means

of achieving a compelling state interest.” McCullen v. Coakley, 573 U.S. 464, 478 (2014).

45 Dave Lawler & Maria Curi, Exclusive: Hegseth gives Anthropic until Friday to back down on

AI safeguards, Axios (Feb. 24, 2026), https://www.axios.com/2026/02/24/anthropic-pentagon-

claude-hegseth-dario.

- 36 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 40 of 48

“When the Government restricts speech, the Government bears the burden of proving the

constitutionality of its actions.” FEC v. Cruz, 596 U.S. 289, 305 (2022).

156. Defendants’ asserted desire to stamp out competing viewpoints about what

Claude can and cannot safely do is not a legitimate interest. See Crime Justice & Am., Inc. v.

Honea, 876 F.3d 966, 973 (9th Cir. 2017) (a government interest is legitimate only if it is

“unrelated to the suppression of expression.”).

157. While the government has a compelling interest in addressing genuine supply

chain risks, Defendants cannot show that the Challenged Actions advance that interest. And to

the extent the government asserts a compelling interest in obtaining AI services without the two

narrow safeguards Anthropic insists upon, the Challenged Actions were not the least-restrictive

means of achieving that interest. The Department had a straightforward and unrestrictive option

that would have fully served that interest: terminate the contract and hire a different developer.

Indeed, Anthropic offered to facilitate a transition to one of its competitor’s systems, and the

Department is reportedly negotiating agreements with one or more frontier AI developers.

158. Defendants’ First Amendment violations have caused Anthropic ongoing and

irreparable harm.

COUNT III

ARTICLE II OF THE U.S. CONSTITUTION; ULTRA VIRES ACTION

(EQUITABLE CAUSE OF ACTION)

(ALL DEFENDANTS)

159. Anthropic incorporates by reference the allegations of the preceding paragraphs.

160. “The ability to sue to enjoin unconstitutional actions by state and federal officers

is the creation of courts of equity, and reflects a long history of judicial review of illegal

executive action, tracing back to England.” Armstrong, 575 U.S. at 327. “When an executive acts

ultra vires, courts are normally available to reestablish the limits on his authority.” Reich, 74

F.3d at 1328. “[I]t remains the responsibility of the judiciary to ensure that the President act[s]

within those limits” that Congress and the Constitution place on him. Am. Forest Res. Council v.

- 37 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 41 of 48

United States, 77 F.4th 787, 797 (D.C. Cir. 2023); accord Murphy Co. v. Biden, 65 F.4th 1122,

1129-31 (9th Cir. 2023).

161. Under longstanding Supreme Court precedent, “[t]he President’s power, if any, to

issue [that] order must stem either from an act of Congress or from the Constitution itself.”

Youngstown Sheet & Tube Co. v. Sawyer, 343 U.S. 579, 585 (1952).

162. The February 27 Presidential Directive purported to order “EVERY Federal

Agency in the United States Government to IMMEDIATELY CEASE all use of Anthropic’s

technology.”

163. The President has no inherent Article II authority for the Presidential Directive.

There is no “executive practice, long pursued to the knowledge of the Congress and never before

questioned,” Youngstown, 343 U.S. at 610 (Frankfurter, J., concurring), of Presidents using their

official position to punish corporations for expressing views on matters of public concern in

negotiations with the government. The “President enjoys no inherent authority,” Learning Res.,

Inc. v. Trump, 2026 WL 477534, at *7 (U.S. Feb. 20, 2026), to force companies to choose

between removing critical use limitations from their products or suffer immediate and

widespread debarment at the hands of the government. No other President has even attempted to

claim such powers.

164. Nor is there any statutory authority for such a directive. Congress has enacted a

comprehensive statutory regime governing federal procurement. This includes statutes in Title 41

of the U.S. Code, as well as those in Title 10, which are specific to the Department. The

government also has promulgated thousands of pages of regulations and individual agency

guidance that comprehensively address how procurement authority is administered. Under this

detailed framework, if the government and a contractor cannot agree on terms for procured

services, the ordinary remedy is for the government not to award a contract or to terminate an

awarded contract for its convenience. See 48 C.F.R. § 49.502. Debarment is not a remedy for

mere contract failure; rather, it is limited to addressing specific “serious . . . irregularities,” may

never be used “for purposes of punishment,” and may only be consummated after providing

robust procedural protections. 48 C.F.R. § 9.402(b); see 48 C.F.R. subpart 9.4.

- 38 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 42 of 48

165. The President’s directive finds no support in this calibrated statutory and

regulatory framework. And even the President cannot “attempt[] to delegate to himself the power

to act arbitrarily.” Anti-Fascist Refugee Committee v. McGrath, 341 U.S. 123, 138 (1951). The

President likewise cannot direct federal agencies to disregard their duly promulgated regulations.

Cf. Nat’l Env’t Dev. Ass’n’s Clean Air Proj. v. EPA, 752 F.3d 999, 1009 (D.C. Cir. 2014) (“[An]

agency is not free to ignore or violate its regulations while they remain in effect.”). The

President’s abrupt directive to cancel Anthropic’s contracts en masse violates these foundational

principles.

166. Finally, the Presidential Directive “possess[es] almost every quality of [an

unlawful] bill[] of attainder.” McGrath, 341 U.S. at 143-44 (Black, J., concurring). It functions

as a “prepared and proclaimed government blacklist[],” punishing Anthropic—and only

Anthropic—without any formal investigation, trial, or even informal process. From the

Founding, such measures have been “forbidden to both national and state governments.” Id. at

144. It cannot be “that the authors of the Constitution, who outlawed the bill of attainder,

inadvertently endowed the executive with [the] power to engage in the same tyrannical practices

that had made the bill such an odious institution.” Id.

167. The President’s ultra vires directive, and any actions by other Defendants

implementing the Presidential Directive, have caused Anthropic ongoing and irreparable harm.

COUNT IV

FIFTH AMENDMENT TO THE U.S. CONSTITUTION (DUE PROCESS)

(EQUITABLE CAUSE OF ACTION; 5 U.S.C. § 702)

(ALL DEFENDANTS)

168. Anthropic incorporates by reference the allegations of the preceding paragraphs.

169. The Fifth Amendment’s Due Process Clause guarantees that “[n]o person shall

. . . be deprived of life, liberty, or property, without due process of law.” U.S. Const. amend.

170. To succeed on its procedural due process claim, Anthropic must show (1) a

deprivation of a protected liberty or property interest; (2) by the government; (3) without the

process that is due under the Fifth Amendment. E.g., Reed v. Goertz, 598 U.S. 230, 236 (2023).

- 39 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 43 of 48

171. The Challenged Actions implicate multiple interests protected by the Due Process

Clause. They impair Anthropic’s liberty interest in its reputation. Wisconsin v. Constantineau,

400 U.S. 433, 437 (1971). They also deprive Anthropic’s property interest in its existing

contracts with the government and private sectors. See Al Haramain Islamic Found. v. U.S.

Dep’t of Treasury, 686 F.3d 965, 973, 979-80 (9th Cir. 2011); Ulrich v. City & Cnty. of San

Francisco, 308 F.3d 968, 976 (9th Cir. 2002) (“‘[I]t has long been settled that a contract can

create a constitutionally protected property interest[.]’”). They purport to (1) terminate

Defendants’ contracts with Anthropic, (2) require many of Anthropic’s largest customers to

terminate their contracts with Anthropic, (3) prohibit Anthropic from participating in federal

contracting, and (4) bar Anthropic from engaging in any future business with any entity that

contracts with the Department.

172. In addition, by purporting to exclude Anthropic from contracting with any federal

agency (apparently for all time), they accomplish a de facto debarment that infringes on

Anthropic’s liberty interest in pursuing its chosen trade. See Trifax Corp. v. District of Columbia,

314 F.3d 641, 643-44 (D.C. Cir. 2003) (“Debarring a corporation from government contract

bidding constitutes a deprivation of liberty that triggers the procedural guarantees of the Due

Process Clause.”); see also Old Dominion Dairy Prods, Inc. v. Sec’y of Def., 631 F.2d 953, 955-

56 (D.C. Cir. 1980); Eng’g v. City & Cnty. of San Francisco, 2011 WL 13153042, at *7 (N.D.

Cal. Feb. 14, 2011).

173. The Challenged Actions imposed these draconian punishments on Anthropic

without any meaningful process. Defendants did not provide Anthropic with any factual findings

remotely supporting the actions taken, much less a meaningful opportunity to challenge them. In

short, the government took these punitive actions “without providing the ‘core requirements’ of

due process: adequate notice and a meaningful hearing.” Jenner & Block LLP v. U.S. Dep't of

Just., 784 F. Supp. 3d 76, 108-09 (D.D.C. 2025) (citation omitted). “[I]f the government must

provide due process before terminating a contractor of its own, surely it must do the same before

blacklisting an entity from all its contractors’ Rolodexes.” Id. at 109.

- 40 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 44 of 48

174. To the extent that a formal process did occur out of public view, it is clear that the

outcome was fatally predetermined by the Department’s retaliatory animus. Prejudgment and

process tainted by animus do not satisfy the requirements of the Due Process Clause.

175. Defendants’ violations of due process have caused Anthropic ongoing and

irreparable harm.

COUNT V

ADMINISTRATIVE PROCEDURE ACT

(5 U.S.C. §§ 558, 706(2))

(ALL AGENCY DEFENDANTS)

176. Anthropic incorporates by reference the allegations of the preceding paragraphs.

177. The APA provides that “[a] sanction may … be imposed or a substantive . . .

order issued [only] within jurisdiction delegated to the agency and as authorized by law.” 5

U.S.C. § 558(b). Thus, the APA prohibits an agency from imposing sanctions or issuing orders

that exceed the scope of authority delegated to it by Congress.

178. After the President issued the Presidential Directive on February 27, numerous

agencies promptly issued sanctions and orders against Anthropic.

179. For example, the Secretarial Order did not only purport “to designate Anthropic a

Supply-Chain Risk to National Security,” it also directed that, “[e]ffective immediately, no

contractor, supplier, or partner that does business with the United States military may conduct

any commercial activity with Anthropic.” The Secretarial Letter issued on March 4 purported to

formalize that final decision.

180. Later on Friday, February 27, 2026, GSA issued an order removing Anthropic

from its Multiple Awards Schedule and USAi.gov. The Multiple Awards Schedule is the federal

government’s primary vehicle for procurement that previously allowed Anthropic to compete for

procurement opportunities at the federal, state, and local level. USAi.gov is a “sandbox” or

- 41 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 45 of 48

centralized platform for federal agencies to test, experiment with, and deploy AI models from

leading providers, including—up to GSA’s action—Anthropic.46

181. Also on February 27, 2026, HHS reportedly took immediate steps to “disabl[e]

enterprise Claude” as a result of the President’s directive, thereby eliminating Anthropic’s ability

to continue to provide its services and compete with other AI providers across HHS’s network.47

182. On March 2, 2026, Treasury Secretary Bessent issued a statement on X that the

Treasury was “terminating all use of Anthropic products . . . within the department” because the

“American people deserve confidence that every tool in government serves the public interest.”48

The same day, the State Department announced that it was “taking immediate steps to implement

the [President’s] directive” and switch “the model powering its in-house chatbot . . . to OpenAI

from Anthropic.”49 The Federal Housing Finance Agency also released statements that it and

mortgage agencies Fannie Mae and Freddie Mac would terminate all use of Anthropic

products.50

183. On information and belief, additional federal agencies are positioned to issue

similar directives and orders.

184. These actions are substantive “orders” within the meaning of 5 U.S.C. § 558(b)

because they are “final disposition[s] . . . of an agency in a matter other than rule making.”

5 U.S.C. § 551(6). These actions also are “sanctions” within the meaning of Section 558(b)

because they impose “limitation[s]” and “other . . . restrictive action[s]” affecting Anthropic’s

46 Press Release, Gen. Servs. Admin., GSA Stands with President Trump on National Security

AI Directive (Feb. 27, 2026), https://www.gsa.gov/about-us/newsroom/news-releases/gsa-stands-

with-president-trump-on-national-security-ai-directive-02272026.

47 Lindsey Wilkinson & Madison Alder, Anthropic faces fallout across federal agencies from

DOD clash, FedScoop (Feb. 27, 2026), https://fedscoop.com/anthropic-claude-dod-federal-

agency-fallout-trump-hegseth/.

48 Scott Bessent (@SecScottBessent), X (Mar. 2, 2026, 10:57 AM ET),

https://x.com/secscottbessent/status/2028499953283117283.

49 Raphael Satter & Courtney Rozen, State Department switches to OpenAI as US agencies start

phasing out Anthropic, Reuters (Mar. 2, 2026), https://www.reuters.com/business/us-treasury-

ending-all-use-anthropic-products-says-bessent-2026-03-02/.

50 Id.

- 42 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 46 of 48

freedom to compete with other AI providers for procurement opportunities and its ability to

protect its reputation as an AI provider serving the public interest. 5 U.S.C. § 551(10).

185. No statute authorizes federal agencies to impose abrupt and en masse orders and

sanctions limiting Anthropic’s ability to compete and impugning Anthropic’s reputation.

186. “Congress could not speak more clearly than it has in the text of the APA: ‘a

sanction may not be imposed or a substantive . . . order issued except within jurisdiction

delegated to the agency and as authorized by law.’” Am. Bus Ass’n v. Slater, 231 F.3d 1, 7 (D.C.

Cir. 2000) (citing 5 U.S.C. § 558(b)). The Challenged Orders of the non-Department Agencies

are “without statutory authorization,” id., and must be set aside under the APA.

187. Defendants’ APA violations have caused Anthropic ongoing and irreparable

harm.

PRAYER FOR RELIEF

For these reasons, Plaintiff respectfully requests an order that:

1. As to the Secretarial Order:

a. Declares the Secretarial Order, and the implementing Secretarial Letter, arbitrary,

capricious, an abuse of discretion, and contrary to law under 5 U.S.C.

§ 706(2)(A);

b. Declares the Secretarial Order, and the implementing Secretarial Letter, contrary

to constitutional right under 5 U.S.C. § 706(2)(B);

c. Declares the Secretarial Order, and the implementing Secretarial Letter, in excess

of statutory jurisdiction, authority, or limitations under 5 U.S.C. § 706(2)(C);

d. Sets aside and vacates the Secretarial Order, and the implementing Secretarial

Letter, in its entirety under 5 U.S.C. § 706(2);

e. Stays the effective date of the Secretarial Order, and the implementing Secretarial

Letter, under 5 U.S.C. § 705 until the conclusion of judicial proceedings in this

action.

- 43 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEFCase 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 47 of 48

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

2. As to the Presidential Directive:

a. Declares that the Presidential Directive exceeds the President’s authority and

violates the First Amendment and Fifth Amendment to the United States

Constitution.

4. As to all of the Challenged Actions:

a. Permanently enjoins Defendants and all their officers, employees, and agents

from implementing, applying, or enforcing the Challenged Actions;

b. Directs Defendants and their agents, employees, and all persons acting under their

direction or control to rescind any and all guidance, directives, or communications

that have issued relating to the implementation or enforcement of the Challenged

Actions, including the Secretarial Letter;

c. Directs Defendants and their agents, employees, and all persons acting under their

direction or control to immediately issue guidance to their officers, staff,

employees, contractors, and agents to disregard the Challenged Actions and any

implementing directives;

d. Awards Plaintiffs their costs and reasonable attorneys’ fees as appropriate; and

e. Grants such further and other relief as this Court deems just and proper.

- 44 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEF1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

Case 3:26-cv-01996-RFL Document 1 Filed 03/09/26 Page 48 of 48

Date: March 9, 2026 KELLY P. DUNBAR (pro hac forthcoming)

kelly.dunbar@wilmerhale.com

JOSHUA A. GELTZER (pro hac

forthcoming)

joshua.geltzer@wilmerhale.com

KEVIN M. LAMB (pro hac forthcoming)

kevin.lamb@wilmerhale.com

SUSAN HENNESSEY (pro hac forthcoming)

susan.hennessey@wilmerhale.com

LAUREN MOXLEY BEATTY (SBN

308333) (application pending)

lauren.beatty@wilmerhale.com

LAURA E. POWELL (pro hac forthcoming)

laura.powell@wilmerhale.com

SONIKA R. DATA (pro hac forthcoming)

sonika.data@wilmerhale.com

WILMER CUTLER PICKERING

HALE AND DORR LLP

2100 Pennsylvania Avenue NW

Washington, DC 20037

Telephone: (202) 663-6000

/s/ Michael J. Mongan

MICHAEL J. MONGAN (SBN 250374)

michael.mongan@wilmerhale.com

WILMER CUTLER PICKERING

HALE AND DORR LLP

50 California Street, Suite 3600

San Francisco, CA 94111

Telephone: (628) 235-1000

EMILY BARNET (pro hac forthcoming)

emily.barnet@wilmerhale.com

WILMER CUTLER PICKERING

HALE AND DORR LLP

7 World Trade Center

250 Greenwich St

New York, NY 10007

Telephone: (212) 230-8800

Attorneys for Plaintiff Anthropic PBC

- 45 - COMPLAINT FOR DECLARATORY

AND INJUNCTIVE RELIEFCase 3:26-cv-01996-RFL Document 1-1 Filed 03/09/26 Page 1 of 2

EXHIBIT 1Case 3:26-cv-01996-RFL Document 1-1 Filed 03/09/26 Page 2 of 2

Truth Details | Truth Social

Truth Details

4809 replies

Donald J. Trump

@realDonaldTrump

THE UNITED STATES OF AMERICA WILL NEVER ALLOW A RADICAL LEFT, WOKE COMPANY TO DICTATE

HOW OUR GREAT MILITARY FIGHTS AND WINS WARS! That decision belongs to YOUR COMMANDER-IN-

CHIEF, and the tremendous leaders I appoint to run our Military.

The Leftwing nut jobs at Anthropic have made a DISASTROUS MISTAKE trying to STRONG-ARM the

Department of War, and force them to obey their Terms of Service instead of our Constitution. Their

selfishness is putting AMERICAN LIVES at risk, our Troops in danger, and our National Security in

JEOPARDY.

Therefore, I am directing EVERY Federal Agency in the United States Government to IMMEDIATELY

CEASE all use of Anthropic’s technology. We don’t need it, we don’t want it, and will not do business with

them again! There will be a Six Month phase out period for Agencies like the Department of War who are

using Anthropic’s products, at various levels. Anthropic better get their act together, and be helpful

during this phase out period, or I will use the Full Power of the Presidency to make them comply, with

major civil and criminal consequences to follow.

WE will decide the fate of our Country — NOT some out-of-control, Radical Left AI company run by

people who have no idea what the real World is all about. Thank you for your attention to this matter.

MAKE AMERICA GREAT AGAIN!

PRESIDENT DONALD J. TRUMP

13.7k ReTruths 56.7k Likes Feb 27, 2026, 12:47 PM

https://truthsocial.com/@realDonaldTrump/posts/116144552969293195Case 3:26-cv-01996-RFL Document 1-2 Filed 03/09/26 Page 1 of 2

EXHIBIT 2Case 3:26-cv-01996-RFL Document 1-2 Filed 03/09/26 Page 2 of 2

Secretary of War Pete Hegseth on X: "This week, Anthropic delivered a master class in arrogance and betrayal as well as a textboo…

Post

Secretary of War Pete Hegseth

@SecWar

This week, Anthropic delivered a master class in arrogance and betrayal as

well as a textbook case of how not to do business with the United States

Government or the Pentagon.

Our position has never wavered and will never waver: the Department of

War must have full, unrestricted access to Anthropic’s models for every

LAWFUL purpose in defense of the Republic.

Instead, @AnthropicAI and its CEO @DarioAmodei, have chosen duplicity.

Cloaked in the sanctimonious rhetoric of “effective altruism,” they have

attempted to strong-arm the United States military into submission - a

cowardly act of corporate virtue-signaling that places Silicon Valley

ideology above American lives.

The Terms of Service of Anthropic’s defective altruism will never outweigh

the safety, the readiness, or the lives of American troops on the battlefield.

Their true objective is unmistakable: to seize veto power over the

operational decisions of the United States military. That is unacceptable.

As President Trump stated on Truth Social, the Commander-in-Chief and

the American people alone will determine the destiny of our armed forces,

not unelected tech executives.

Anthropic’s stance is fundamentally incompatible with American principles.

Their relationship with the United States Armed Forces and the Federal

Government has therefore been permanently altered.

In conjunction with the President's directive for the Federal Government to

cease all use of Anthropic's technology, I am directing the Department of

War to designate Anthropic a Supply-Chain Risk to National Security.

Effective immediately, no contractor, supplier, or partner that does business

with the United States military may conduct any commercial activity with

Anthropic. Anthropic will continue to provide the Department of War its

services for a period of no more than six months to allow for a seamless

transition to a better and more patriotic service.

America’s warfighters will never be held hostage by the ideological whims

of Big Tech. This decision is final.

2:14 PM · Feb 27, 2026 · 12.8M Views

https://x.com/SecWar/status/2027507717469049070Case 3:26-cv-01996-RFL Document 1-3 Filed 03/09/26 Page 1 of 5

EXHIBIT 3Case 3:26-cv-01996-RFL Document 1-3 Filed 03/09/26 Page 2 of 5

SECRETARY OF WAR

1000 DEFENSE PENTAGON

WASHINGTON, DC 20301-1000

MAR - 3 2026

Mr. Dario Amodei

Chief Executive Officer

Anthropic, PBC

548 Market Street

San Francisco, CA 94104

Dear Anthropic, PBC Executive Leadership:

This letter provides notice to Anthropic, Public Benefit Corporation (PBC), and its

subordinate, subsidiaries, or affiliated offices or entities, doing business under various

names, and all subsidiaries, successors, or assigns thereof ("Covered Entity") that, pursuant

to title 10, United States Code (U.S.C.), section 3252 ("Section 3252"), the Department of

War (DoW) has determined that (i) the use of the Covered Entity's products or services in

Do W covered systems I presents a supply chain risk and that the use of the Section 3252

authority to carry out covered procurement actions2 is necessary to protect national security

by reducing supply chain risk, and (ii) less intrusive measures are not reasonably available

to reduce such supply chain risk.

Scope of Authorized Covered Procurement Actions

This Determination is necessary to reduce supply chain risk and applies to the

Covered Entity, Covered Products or Services, Covered Procurements, and Covered

Procurement Actions as follows:

• Covered Entity: Anthropic, PBC, and its subordinate, subsidiaries, or affiliated

offices or entities, doing business under various names, and all subsidiaries,

successors, or assigns thereof.

• Covered Products or Services: All of the Covered Entity's products or services that

meet the definition of a "covered item of supply" or that would be procured as part of

a "covered system," as those terms are defined at 48 C.F.R. § 239.7301 , whether

acquired as a product or service. This includes all of the Covered Entity's products or

services offered by the Covered Entity that become available for procurement.

• Covered Procurements: All DoW procurements for which 48 C.F.R. Subpart

239.73 is applicable, in accordance with 48 C.F.R. § 239.7302.

• Covered Procurement Actions: All actions authorized in accordance with 48

C.F.R. § 239.7305.

1 10 U. S.C. § 3252(d)(5)

2 IO U.S.C. § 3252(d)(2)Case 3:26-cv-01996-RFL Document 1-3 Filed 03/09/26 Page 3 of 5

Effective Date

This Determination is effective immediately and shall remain in effect until modified

or terminated in writing by the Section 3252 Authorized Official.

Request for Reconsideration

If the Covered Entity wishes to request that the DoW reconsider this Determination,

the Covered Entity must submit in writing to the undersigned within 30 days of receipt of

this letter, notice of such request for reconsideration. For additional information,

requirements, and procedures governing such request, see the enclosed Requirements and

Procedures for Requesting Reconsideration of a Section 3252 Determination.

Sincerely,

Enclosure:

As stated

2UNCLASSIFIED

Case 3:26-cv-01996-RFL Document 1-3 Filed 03/09/26 Page 4 of 5

ATTACHMENT 1

Requirements and Procedures for Requesting Reconsideration of a Section 3252

Determination

The following requirements and procedures govern a Covered Entity's request for

reconsideration of a Section 3252 determination:

1. Notice of Request for Reconsideration: Within thirty (30) days of receiving the Section

3252 Authorized Official's letter notifying the Covered Entity of the Section 3252

determination, the Covered Entity may submit in writing to the Section 3252 Authorized

Official notice of the Covered Entity's request for reconsideration. Such notice should

identify the specific relief or remedy being requested ( e.g., specific modifications to, or

termination in whole or in part, of any elements of the determination). All notices and written

information must be submitted to osd.mc-alex.ousd-a-s.mbx.1 0-usc-section-3252-

determinations@mail .mi 1.

2. Opportunity to Submit and Present Information: If the Covered Entity submits a timely

notice of request for consideration, the Covered Entity will be afforded an additional thirty

(30) calendar days (from the date the Section 3252 Authorized Official received such timely

notice) to submit in writing, and to appear and present, additional information and arguments

in support of such request for reconsideration. The Covered Entity must either send, or make

arrangements to appear and present, the information to representatives of the Section 3252

Authorized Official within the thirty (30) day period. The Section 3252 Authorized Official

may extend the time to appear and submit documentary evidence upon written request by the

Covered Entity.

3. Flexible Procedures: The Section 3252 Authorized Official may use flexible procedures to

allow the Covered Entity to submit and present information in support of the request for

reconsideration. In so doing, the Section 3252 Authorized Official is not required to follow

formal rules of evidence or procedures in creating an official record of the request for

reconsideration and the Official's disposition of that and request.

4. Content of Submissions/Presentations: When submitting and presenting information in

support of a request for reconsideration, the Covered Entity should, to the maximum extent

practicable, identify specific facts that contradict statements contained in the Section 3252

Covered Entity notification, and provide detailed rationale for any arguments in support of

the request and the remedy or relief being requested. A general denial is insufficient to

support reconsideration of a Section 3252 determination.

5. Appearing and Presenting Information: An appearance and presentation is an informal

meeting that is non-adversarial in nature. When electing to appear and present information,

the representative(s) of the Covered Entity may choose to appear with counsel. Any

information to be presented should be provided in written form at least 5 working days in

advance of the presentation. Usually, all matters in opposition should be presented in a single

proceeding. Any information not submitted in advance, but provided orally during an

appearance, must also be submitted in writing after the appearance for the information to be

UNCLASSIFIEDUNCLASSIFIED

Case 3:26-cv-01996-RFL Document 1-3 Filed 03/09/26 Page 5 of 5

6. considered. The representative(s) of the Section 3252 Authorized Official, and/or other

agency representatives, may ask questions of the Covered Entity's representative(s) making

the presentation. Federal rules of evidence do not govern the appearance and presentation.

Notice Regarding False Statements: Any material information submitted in response to this

action will be considered a statement or representation to a government official concerning a

matter within the jurisdiction of the executive branch of the government. To that end, please

note that an individual making any materially false, fictitious, or fraudulent statement or

representation to a Government official may be subject to prosecution under 18 U.S.C. §

1001.

UNCLASSIFIED