The Information Machine

Pentagon and Microsoft Pulling Back from Anthropic Claude · history

Version 4

2026-05-25 05:38 UTC · 155 items

What

Anthropic's bid to shed the Pentagon's 'Supply Chain Risk' designation has failed in D.C. courts at both the district and circuit levels [10][13][14], while the Pentagon is separately appealing the California court order that had partially protected Anthropic's federal contracting [18]. The executive branch conflict is now documented, not merely inferred: the White House cleared Anthropic's classified NSA deal over direct Pentagon objection [24], as part of a $9 billion spy-agency AI spending push [28][26]. A bipartisan coalition of 149 former federal and state judges filed an amicus brief backing Anthropic's suit against the Department of Defense [19]. Microsoft's cancellation of internal Claude Code licenses [32][33] continues as a separate competitive pressure.

Why it matters

The White House overriding Pentagon objections to award Anthropic an intelligence contract [24] transforms what began as a vendor dispute into a live contest over which part of the executive branch controls AI procurement standards — an intra-government conflict that procurement law has rarely had to resolve at this level. The 149 former judges' amicus brief [19] raises the legal stakes further, signaling that the methodology of the Pentagon's designation may carry systemic procedural or constitutional implications beyond this single company.

Open questions

  • Can the White House legally advance the Anthropic-NSA deal over the Pentagon's Supply Chain Risk designation [24] without formally lifting the classification — and if two executive agencies can simply operate under contradictory vendor policies, what does that mean for the coherence of federal procurement law?

  • If the Pentagon prevails in appealing the California court order [18], does Anthropic lose its only remaining judicial protection against broad federal contracting restrictions, leaving only the D.C. litigation where it has lost at both levels?

  • The 149 former judges' amicus brief [19] is an unusually prominent show of cross-partisan legal-community support — has it influenced the D.C. appellate court's reasoning, or does it function primarily as political pressure with limited doctrinal weight?

  • With the White House committing $9 billion for spy-agency AI [28] alongside a Pentagon classified roster that explicitly excludes Anthropic [7], how are these parallel budgets allocated — and does intelligence community revenue neutralize the DoD exclusion's commercial impact?

Narrative

The conflict between Anthropic and the U.S. government traces to a blunt ultimatum over AI deployment in military contexts. In late February 2026, Anthropic CEO Dario Amodei publicly refused Pentagon demands to modify the company's AI safety policies and permit Claude's use in autonomous weapons development and mass surveillance [1][2][3]. The refusal prompted Trump to order the federal government to stop using Anthropic's tools [4]. The Department of Defense — operating as the 'Department of War' under the Trump administration, as documented in federal court filings and confirmed by the agency's own war.gov domain [5][6][7] — responded by designating Anthropic a 'Supply Chain Risk,' a national security procurement classification that bars a company from classified government contracts. The case, Anthropic PBC v. United States Department of War, became the most consequential AI procurement dispute yet litigated in U.S. federal court.

The legal battle produced a split outcome across jurisdictions, with multiple rulings now confirming that Anthropic's position in the controlling D.C. jurisdiction is precarious. A California federal court in May 2026 blocked broad restrictions on Anthropic's federal contracting access [8], but D.C. courts ruled the opposite way at both levels: a D.C. district court denied Anthropic's motion to lift the designation in April [9][10][11], and the D.C. Circuit subsequently upheld the Pentagon's classification [12][13][14]. Legal commentary firm Vital Law captured the asymmetry: 'After Win in California, Anthropic Loses to DoD in D.C. Court' [15]. Legal analysis firms including Mayer Brown and Fluet Law published guidance for government contractors navigating the designation's practical effects [16][17]. The Pentagon has since moved to appeal the California order that had offered Anthropic's only remaining judicial protection [18]. A bipartisan coalition of 149 former federal and state judges filed an amicus brief in March 2026 supporting Anthropic's suit [19] — the most prominent show of legal-community backing yet — though as Politico sources warned as early as March 27, California court wins should not be read as resolving the D.C. threat [20].

The clearest signal of intra-executive conflict has now become explicit. The Department of War expanded its classified AI vendor roster to seven companies according to The Guardian [21] and its own press release [7], or eight according to DefenseScoop [22] — in either case a list that explicitly excludes Anthropic — while evaluating rival models with 25 power users as direct Claude replacements [23]. Yet the White House reportedly cleared a classified Anthropic-NSA deal over direct Pentagon objection [24], with U.S. officials and Anthropic finalizing a $125 million arrangement to give spy agencies access to Claude [25][26][27] — framed within a broader $9 billion White House commitment to advance spy-agency AI capabilities [28]. Trump himself has said a DoD deal with Anthropic is 'possible' [29], and the intelligence community deal was reportedly pushed through despite, not alongside, the military's standing objection. Two parts of the same executive branch are simultaneously pursuing contradictory procurement policies toward the same vendor.

Anthropologic's commercial trajectory suggests the military exclusion has not materially impaired the company's growth: it is on track for $10.9 billion in Q2 2026 revenue and projected to reach its first profitable quarter [30], with Reuters reporting a goal to nearly triple annualized revenue in 2026 [31]. Microsoft separately canceled internal Claude Code developer licenses and redirected engineers to GitHub Copilot [32][33][34], a move attributed to financial consolidation around Microsoft's own tooling rather than capability concerns [32]. The Brookings Institution frames the Pentagon dispute as a potential inflection point for responsible AI in national security: if Anthropic cannot sustain ethical limits against determined executive pressure, it raises the question of whether responsible AI commitments can survive military procurement demands industry-wide [35]. The Atlantic takes the opposing view, arguing the stand may be strategically and ethically vindicating [36][37]. International institutions including SIPRI and UNIDIR have published parallel frameworks on responsible military AI procurement norms [38][39][40], providing global context for the standards questions the Anthropic-DoD standoff is forcing into public view.

Timeline

  • 2026-02: Defense One and Scientific American report that replacing Anthropic's AI tools at the Pentagon would take months, and sources warn the transition won't be straightforward. [48][49]
  • 2026-02-26: The Washington Post and The Verge report that Anthropic has rejected Pentagon terms for lethal and autonomous weapons use of Claude. BBC covers CEO Dario Amodei's public refusal to drop AI safeguards for the Pentagon. [1][2][3]
  • 2026-02: Trump orders the U.S. government to stop using Anthropic's AI tools after Anthropic refuses to permit use for autonomous weapons and mass surveillance. [4]
  • 2026-03-10: DefenseScoop reports the Pentagon launches a Google Gemini-based 'GenAI Agent Designer' allowing DoD employees to create custom AI assistants. [50]
  • 2026-03: Federal News Network reports the DoD is confident it can replace Claude within six months. Anthropic announces a $200M DoD agreement for responsible AI in defense operations. Anthropic challenges the Department of War's Supply Chain Risk designation in court. The Atlantic publishes analysis arguing Anthropic's ethical stand may be paying off. [51][46][44][36]
  • 2026-03-18: A bipartisan coalition of 149 former federal and state judges files an amicus brief supporting Anthropic's suit against the Department of Defense. [19]
  • 2026-03-27: Politico reports that despite a court win in California, lawyers and lobbyists warn Anthropic remains 'in trouble' and the win should not be read as decisive. [20]
  • 2026-04-08: A federal district court in D.C. denies Anthropic's motion to lift the Department of War's Supply Chain Risk designation; CNBC, Axios, and KQED confirm the ruling. [9][10][11][13]
  • 2026-04-21: Trump says an Anthropic deal for Department of Defense use is 'possible.' [29]
  • 2026-04-28: Google signs an AI deal with the Pentagon. [52]
  • 2026-04-29: Axios reports Congress is stalling on military AI legislation as the Google-Pentagon AI deal moves forward. [53]
  • 2026-05-01: The Pentagon strikes AI deals with seven to eight major tech companies for classified network AI deployment, explicitly excluding Anthropic. The Department of War publishes an official press release on the classified networks AI agreements. [22][54][55][21][7]
  • 2026-05-19: D.C. Circuit judges question the legality of the DoD's move to bar Anthropic from government contracts during appellate proceedings. [56][57]
  • 2026-05-20: CNBC reports Anthropic is on track for $10.9 billion in Q2 revenue, projecting its first profitable quarter. [30]
  • 2026-05-21: Bloomberg reports the Pentagon is testing rival AI models with 25 power users to replace Anthropic's Claude. A California federal court blocks the Trump administration from broadly restricting Anthropic's federal contracts. Multiple outlets confirm Microsoft has canceled internal Claude Code developer licenses and is redirecting employees to GitHub Copilot. [23][8][58][32][33][34]
  • 2026-05-22: The D.C. Circuit upholds the Pentagon's Supply Chain Risk designation against Anthropic. The White House approves $9 billion for spy agencies to advance AI capabilities. Reports emerge of a classified NSA contract with Anthropic. [12][15][59][28]
  • 2026-05: The Pentagon appeals the California federal court order that had blocked broad restrictions on Anthropic's federal contracting. [18]
  • 2026-05-24: Reports from The Information and others confirm the White House and Anthropic are finalizing a $125 million deal for U.S. spy agencies to access Claude. AI Weekly reports the White House cleared the deal over direct Pentagon objection. [25][60][61][27][26][24]

Perspectives

U.S. Department of War (formerly Pentagon/DoD)

Demanded Anthropic modify usage terms to allow weapons and surveillance applications; after refusal, designated Anthropic a Supply Chain Risk and sought to bar it from government contracting; prevailed in D.C. courts at both district and circuit levels; expanded classified AI work to seven or eight major companies that explicitly exclude Anthropic, as confirmed in the DoW's own official press release; now appealing the California court order that had partially protected Anthropic's contracting access.

Evolution: The Pentagon's appeal of the California ruling [17852] is new — it attempts to close the one judicial gap remaining in Anthropic's favor. The DoW's primary-source press release [18433] now provides direct official confirmation of Anthropic's exclusion from the classified AI vendor roster.

[41][42][43][4][44][9][12][22][7][18]

Anthropic

Refused the Pentagon's weapons and surveillance ultimatum; failed to lift the Supply Chain Risk designation in D.C. courts; retains a partial California judicial shield now under Pentagon appeal; holds a $200M DoD agreement alongside the reportedly finalizing $125M White House intelligence community deal; CEO Dario Amodei has spoken publicly about the specific Pentagon uses he refuses to allow.

Evolution: Amodei's public statements are now documented across mainstream outlets including BBC [15755] and social media [18429], elevating the profile and directness of Anthropic's position. The amicus brief from 149 former judges [18430] represents significant external legal-community support that was not present in prior coverage.

[2][1][3][45][43][46][25][20][15][19]

Federal judiciary

Split across jurisdictions: a California federal court blocked broad federal contracting restrictions on Anthropic; D.C. district and circuit courts upheld the Pentagon's Supply Chain Risk designation; the Pentagon is now appealing the California order, which could eliminate the only judicial protection Anthropic holds outside D.C.

Evolution: The Pentagon's appeal of the California order [17852] adds a new procedural front that, if successful, would leave Anthropic with no active judicial protection against the designation's effects.

[8][9][12][15][10][13][14][18]

Trump administration / White House

Trump ordered the government-wide ban on Anthropic tools after the weapons-and-surveillance refusal, and has also said a DoD deal is 'possible'; the White House separately cleared an Anthropic-NSA deal over direct Pentagon objection and is committing $9 billion in spy-agency AI spending, demonstrating a documented internal contradiction between its military and intelligence procurement postures toward Anthropic.

Evolution: The White House clearing the NSA deal over explicit Pentagon objection [18431] is the most concrete evidence yet of intra-executive division — previously this was inferred from parallel deal reports; now it is documented as an override of the military's position.

[4][29][25][28][26][24]

Bipartisan coalition of 149 former federal and state judges

Filed an amicus brief supporting Anthropic's suit against the Department of Defense, signaling that a broad, cross-partisan legal community views the government's Supply Chain Risk designation tactics as raising significant procedural or constitutional concerns.

Evolution: First appearance in synthesis — this coalition was not present in prior coverage and represents a novel form of external pressure on the litigation.

[19]

The Atlantic

Frames Anthropic's refusal of the Pentagon's weapons and surveillance ultimatum as an ethical stand that may ultimately prove strategically advantageous.

Evolution: Consistent with prior synthesis.

[36][37]

Brookings Institution

Poses the dispute as a potential inflection point for responsible AI in national security: if Anthropic cannot sustain ethical restrictions against the Pentagon, it raises the question of whether responsible AI commitments can survive military procurement demands industry-wide.

Evolution: Consistent with prior synthesis.

[35]

Microsoft

Canceled internal Claude Code developer licenses and directed employees to GitHub Copilot CLI; Claude remains accessible to employees through Copilot CLI and continues to power features inside Microsoft 365.

Evolution: Consistent with prior synthesis.

[47][32][33][34]

Legal analyst community (Mayer Brown, Fluet Law)

Published detailed guidance on the Supply Chain Risk designation's practical effects for government contractors working with AI, framing it as a landmark procurement law case with broad industry implications beyond Anthropic itself.

Evolution: First appearance in synthesis — legal industry commentary has emerged as a distinct analytical voice as the litigation matured.

[16][17]

Tensions

  • Department of War vs. Anthropic on autonomous weapons use: the DoW demands Claude be deployable for weapons development and mass surveillance; Anthropic's ethical policies prohibit this and CEO Dario Amodei has publicly refused to change them, triggering the Supply Chain Risk designation and all subsequent legal and procurement consequences. [2][1][3][45][4][43][44]
  • D.C. courts vs. California courts on Anthropic's legal status: Anthropic won a preliminary ruling in California blocking broad contracting restrictions, but the D.C. Circuit and D.C. district court upheld the Pentagon's supply chain risk designation — and the Pentagon is now appealing the California ruling, attempting to close the one judicial gap remaining in Anthropic's favor. [8][9][12][15][10][13][14][18]
  • Department of War military exclusion vs. White House intelligence community inclusion: the DoW bars Anthropic via the Supply Chain Risk designation and expanded AI contracts to companies that explicitly exclude Anthropic [18433], while the White House cleared a $125M Anthropic-NSA deal over direct Pentagon objection [18431] — two parts of the same executive branch pursuing directly contradictory procurement policies toward the same vendor. [22][7][21][25][26][24][28]
  • The Atlantic vs. Brookings on the implications of Anthropic's ethical stand: The Atlantic frames the refusal as potentially strategic and vindicating for responsible AI development; Brookings asks whether the feud instead signals the end of responsible AI commitments surviving contact with military procurement — opposite readings of the same standoff's meaning for the industry. [36][37][35]
  • Microsoft motive — financial discipline vs. strategic lock-in: Windows Central frames the Claude Code license cancellation as 'likely financial'; The Decoder frames it as Microsoft redirecting developers to its own tooling. Both can be true, but the emphasis carries different implications for Anthropic's relationship with Microsoft's Copilot ecosystem going forward. [32][33]

Sources

  1. [1] Anthropic rejects Pentagon terms for lethal use of its chatbot Claude — reactive:anthropic-enterprise-losses
  2. [2] Anthropic refuses Pentagon's new terms, standing firm on lethal ... — reactive:anthropic-enterprise-losses
  3. [3] Anthropic boss rejects Pentagon demand to drop AI safeguards — reactive:anthropic-ai-values-widening
  4. [4] Anthropic wanted the Pentagon to agree not to use its AI for autonomous weapons and mass surveillance. So Trump ordered the government to stop using it altogether. — reactive:anthropic-enterprise-losses
  5. [5] Anthropic PBC v. United States Department of War 26-01049 (D.C. Cir.) — reactive:anthropic-enterprise-losses
  6. [6] Anthropic PBC v. United States Department of War, 26-1049 – CourtListener.com — reactive:anthropic-enterprise-losses
  7. [7] Classified Networks AI Agreements - U.S. Department of War — reactive:anthropic-enterprise-losses
  8. [8] Judge blocks Trump administration from limiting Anthropic's contracts with federal government — reactive:anthropic-enterprise-losses
  9. [9] Federal Court Denies Anthropic's Motion to Lift 'Supply ... — reactive:anthropic-enterprise-losses
  10. [10] Anthropic loses appeals court bid to temporarily block DOD ruling — reactive:anthropic-ai-values-widening
  11. [11] Anthropic loses bid to block Pentagon blacklisting in DC court - Axios — reactive:anthropic-enterprise-losses
  12. [12] Federal Appeals Court Upholds Pentagon's Supply-Chain ... — reactive:anthropic-enterprise-losses
  13. [13] Anthropic’s Bid to Lift ‘Supply Chain Risk’ Label Suffers Setback in US Appeals Court | KQED — reactive:anthropic-enterprise-losses
  14. [14] Appeals Court Denies Anthropic's Bid to Shed 'Supply Chain Risk ... — reactive:anthropic-enterprise-losses
  15. [15] After Win in California, Anthropic Loses to DoD in D.C. Court — reactive:anthropic-enterprise-losses
  16. [16] Anthropic Supply Chain Risk Designation Takes Effect - Mayer Brown — reactive:anthropic-enterprise-losses
  17. [17] The Department's War with Anthropic: Litigation Update - Fluet Law — reactive:anthropic-enterprise-losses
  18. [18] Pentagon appealing order to remove Anthropic 'supply chain risk' label — reactive:anthropic-enterprise-losses
  19. [19] BIPARTISAN COALITION OF 149 FORMER FEDERAL AND STATE JUDGES FILES BRIEF SUPPORTING ANTHROPIC’S SUIT AGAINST THE DEPARTMENT OF DEFENSE — reactive:anthropic-enterprise-losses
  20. [20] 'Premature': Anthropic still in trouble despite court win ... — reactive:anthropic-enterprise-losses
  21. [21] Pentagon inks deals with seven AI companies for classified military work — reactive:openai-microsoft-partnership-amendment
  22. [22] DOD expands its classified AI work with 8 companies — excluding Anthropic — amid ongoing dispute | DefenseScoop — reactive:anthropic-enterprise-losses
  23. [23] Pentagon Tests Rival AI Models in Race to Replace ... — reactive:anthropic-enterprise-losses
  24. [24] White House Clears Anthropic NSA Deal Over Pentagon Objection | AI Weekly — reactive:anthropic-enterprise-losses
  25. [25] 🤯 ANTHROPIC just secured a $125M White House Intel deal. — reactive:anthropic-enterprise-losses (2026-05-24)
  26. [26] White House, Anthropic Near Deal For Spy Agencies to Use AI — reactive:anthropic-enterprise-losses
  27. [27] White House Nears Deal with Anthropic for AI Use in Intelligence — reactive:anthropic-enterprise-losses
  28. [28] White House Approves $9 Billion for Spy Agencies to Catch Up on AI - GV Wire — reactive:anthropic-enterprise-losses
  29. [29] Trump says Anthropic deal is 'possible' for Department of Defense use — reactive:anthropic-enterprise-losses
  30. [30] Anthropic set to hit $10.9 billion in revenue in Q2, source says - CNBC — reactive:anthropic-enterprise-losses
  31. [31] Exclusive: Anthropic aims to nearly triple annualized revenue in 2026, sources say — reactive:anthropic-enterprise-losses
  32. [32] Microsoft is ditching Claude Code for Copilot CLI - Windows Central — reactive:anthropic-enterprise-losses
  33. [33] Microsoft pulls Claude Code licenses and pushes developers back ... — reactive:anthropic-enterprise-losses
  34. [34] Microsoft starts canceling Claude Code licenses — reactive:anthropic-enterprise-losses
  35. [35] Does the Anthropic–Pentagon feud mean the end of responsible AI? — reactive:anthropic-enterprise-losses
  36. [36] Anthropic’s Ethical Stand Could Be Paying Off - The Atlantic — reactive:anthropic-enterprise-losses
  37. [37] Donald Trump Declares War on Anthropic - The Atlantic — reactive:anthropic-enterprise-losses
  38. [38] [PDF] Responsible Procurement of Military Artificial Intelligence - SIPRI — reactive:anthropic-enterprise-losses
  39. [39] Framework of Responsible Industry Behaviour for AI in the Military Domain → UNIDIR — reactive:anthropic-enterprise-losses
  40. [40] Responsible behaviour in military AI starts with responsible procurement | SIPRI — reactive:anthropic-enterprise-losses
  41. [41] The Pentagon is trying to force Anthropic company to break the law ... — reactive:anthropic-enterprise-losses
  42. [42] Pentagon Threatens to End Anthropic Work in Feud Over AI Terms — reactive:anthropic-enterprise-losses
  43. [43] Anthropic AI rejects Pentagon's weapons & surveillance ultimatum — reactive:anthropic-enterprise-losses
  44. [44] Anthropic Challenges DoW’s Supply Chain Risk Designation — reactive:anthropic-enterprise-losses
  45. [45] Anthropic's CEO on the 2 things he refuses to let the Pentagon use ... — reactive:anthropic-enterprise-losses
  46. [46] Anthropic awarded $200M DOD agreement for AI capabilities \ Anthropic — reactive:anthropic-enterprise-losses
  47. [47] Microsoft Cancels Internal Claude Code Licenses, Pushes Copilot ... — reactive:anthropic-enterprise-losses
  48. [48] It would take the Pentagon months to replace Anthropic’s AI tools: sources - Defense One — reactive:anthropic-enterprise-losses
  49. [49] Why replacing Anthropic at the Pentagon could take months | Scientific American — reactive:anthropic-enterprise-losses
  50. [50] Pentagon says employees can create their own ‘custom AI assistants’ with new tech | DefenseScoop — reactive:anthropic-enterprise-losses
  51. [51] DoD ‘confident’ it can replace Anthropic’s Claude within six months, but some warn transition won’t be easy | Federal News Network — reactive:anthropic-enterprise-losses
  52. [52] Google Signs A.I. Deal With the Pentagon - The New York Times — reactive:openai-financial-strategy
  53. [53] Congress stalls on military AI as Google and the Pentagon strike deal — reactive:anthropic-enterprise-losses
  54. [54] Pentagon Makes Deals With A.I. Companies to Expand Classified ... — reactive:sweep
  55. [55] Pentagon strikes deals with 7 Big Tech companies after shunning ... — reactive:sweep
  56. [56] Potential splits emerged between D.C. Circuit judges questioning the legality of the DOD's move to bar Anthropic from go... — reactive:anthropic-enterprise-losses (2026-05-20)
  57. [57] Potential splits emerged between D.C. Circuit judges questioning the legality of the DOD's move to bar Anthropic from go... — reactive:anthropic-enterprise-losses (2026-05-19)
  58. [58] 🇺🇸The Pentagon is racing to replace Claude and Anthropic is fighting back. — reactive:anthropic-enterprise-losses (2026-05-21)
  59. [59] More: U.S. officials said Anthropic and the government are finalizing a classified contract that would allow the NSA to ... — reactive:anthropic-enterprise-losses (2026-05-22)
  60. [60] @WatcherGuru Claude's about to get top-secret clearance — reactive:anthropic-enterprise-losses (2026-05-23)
  61. [61] *WHITE HOUSE AND ANTHROPIC NEAR DEAL FOR US SPY AGENICES TO USE ITS AI TOOLS: NYT — reactive:anthropic-enterprise-losses (2026-05-23)