The Civic Node

Anthropic

6 min read

Overview

Anthropic is a US AI safety company founded in 2021, creator of the Claude AI assistant. In early 2026, it was blacklisted by the US Department of Defense as a “national-security supply-chain risk” after refusing to allow Claude to be used for military surveillance or autonomous weapons. The case represents the first major direct confrontation between a large AI company and the US executive branch over military AI use.

Key Facts

  • Creator of Claude AI chatbot; Claude Code is a key enterprise revenue driver
  • Refused to allow Claude to be used for US military surveillance or autonomous weapons Britain woos Anthropic expansion after US defence clash
  • Trump directed federal agencies to cease Anthropic use: February 27, 2026
  • Pentagon designated Anthropic a supply-chain risk: March 5, 2026
  • US judge temporarily blocked the blacklisting: March 26, 2026; second lawsuit pending
  • Britain (Keir Starmer’s government) actively recruiting Anthropic to expand to UK; proposals include London office + dual stock listing
  • CEO: Dario Amodei; UK visit planned for late May 2026
  • Valuation: $183B (Series F, $13B raised, October 2025)
  • Revenue: ~$7B ARR as of October 2025; projected $9B ARR by year-end 2025; 2026 target: up to $26B
  • Actual booked revenue as of July 2025: ~$1.5B (per Zitron analysis) — significantly below ARR figures
  • Burns $5B+/year; not profitable
  • Holds $200M DoD contract and GSA App Store access despite political conflict with Trump administration
  • Offered Claude to all government customers for $1/year (2025)
  • Politically opposed to Trump: opposed 10-year state AI regulation moratorium in “Big Beautiful Bill”; endorsed California SB 53 (AI transparency); hired senior Biden-era officials for government relations
  • Dario Amodei compared Trump to a “feudal warlord” during 2024 election; publicly supported Kamala Harris; excluded from White House dinners attended by OpenAI, Meta, and Nvidia leaders
  • Co-founder Jack Clark’s essay “Technological Optimism and Appropriate Fear” (October 2025) triggered David Sacks’s “regulatory capture through fear-mongering” accusation
  • April 21, 2026 — Claude Mythos breach investigation: Anthropic confirmed it is investigating a Bloomberg report that users in a private forum gained unauthorised access to Claude Mythos (its gated cyber-security model) “through one of our third-party vendor environments” — most likely vendor-permission misuse rather than an external hack. No evidence Anthropic’s own systems affected; no confirmation the model is in malicious hands. Parallels with OpenAI’s GPT 5.4 Cyber. Claude Mythos Unauthorised Access — BBC
  • May 4, 2026 — Anthropic break drives White House toward pre-release AI review: NYT reports Trump administration considering an executive order establishing a working group on AI and a pre-release government review for new AI models, modeled on a UK multi-agency safety review. White House officials have met with Anthropic, Google, OpenAI executives. Reporting attributes the policy reversal (from Trump’s July 2025 deregulatory posture) to the post-February Anthropic break and to White House concern over an AI-enabled cyberattack risk specifically tied to Claude Mythos. Federal appeals court (early May 2026) refused to stay the Pentagon supply-chain risk designation while Anthropic’s lawsuit proceeds. Despite the dispute, Jack Clark said Anthropic is in talks with the federal government about Claude Mythos. Trump May Review AI Models Before Release — Forbes - 2026-05-04

Newsletter Relevance

Power: Anthropic is the clearest current example of a private technology company refusing state demands to militarize its product and being punished with regulatory destruction. The Pentagon’s “supply-chain risk” designation is a new kind of weapon — regulatory rather than legal or military.

DePIN bridge: Anthropic’s situation is an AI-domain preview of pressures DePIN networks will face. If a DePIN energy or communications network becomes strategically important, the state will want to co-opt or control it. Anthropic’s refusal and the resulting blacklist is a model for what that confrontation looks like.

AI Sovereignty: The UK’s aggressive recruitment of Anthropic shows how geopolitical competition for AI infrastructure works in practice — US pressure creates openings for allies/rivals.

Connections

Source Appearances

Open Questions

  • What specifically did the DoD ask Anthropic to enable? Answered (partially): Amodei’s statement specifies the DoD demanded “any lawful use” and removal of safeguards against mass domestic surveillance and fully autonomous weapons. The American Progress piece adds that Anthropic feared tools being used to analyze “bulk data collected from Americans” and “unclassified commercial data.”
  • What is the status of the two lawsuits? What legal theory is Anthropic using?
  • Does Anthropic have meaningful DoD/government revenue at stake, or is this about precedent?
  • What does a “dual stock listing” in the UK actually offer Anthropic?
  • Are other AI companies (OpenAI, Google DeepMind) complying with DoD requests that Anthropic refused? Answered (partially): OpenAI signed a DoD contract within hours of the Anthropic designation, claiming its deal has “more guardrails” including shared red lines on mass surveillance and autonomous weapons. But the American Progress article asks the key follow-up: why would the government honor OpenAI’s restrictions after trying to destroy Anthropic for similar ones?

Graph View

Backlinks