ANTHROPIC

Institutional Archetype

THE BRAKE THAT SELLS CARS — Anthropic is the frontier lab that argues, in writing and with money, that governments should be able to slow or stop frontier labs, its own included. It builds a top-ranked model, and it funds the policy campaign for rules that would bind the field. The two halves of the position are the archetype: the company most insistent that the frontier is dangerous is also one of the two companies at its edge. Its critics read the brake as a moat. Its own record reads the brake as the point. The state, in the same year, both blacklisted it as a supply-chain risk and, per a federal judge, kept working with its newest model on cybersecurity. This profile scores the position (the lab whose product is also the case for regulating labs), not the people in it.

Mandate & Origin

Anthropic announced a $124 million Series A on 28 May 2021, describing itself as “an AI safety and research company” led by siblings Dario Amodei (CEO) and Daniela Amodei (President) (Anthropic, 28 May 2021). Its founding team came out of OpenAI; for the founders and the departure, see Dario Amodei and Jack Clark. Its product line is Claude. On Artificial Analysis’s intelligence index as reported in September 2026, Claude Opus 5.5 leads with a score of 58 (AI Magazine).

The stated mandate, in 2026. The company’s July 2026 policy statement: “Governments should be able to verify companies’ safety claims, enforce safe practices through civil penalties, and ultimately have a way to slow or block the deployment of AI models that pose a serious risk of catastrophic harm. Frontier AI developers should have to test models that pose catastrophic risk, be transparent to the public about their findings, submit them to independent evaluation, and maintain a robust security program.” It calls its Advanced AI Framework “the strongest policy proposal from any frontier lab or policymaker to date” (Anthropic, 21 Jul 2026).

On open weights. “Anthropic has never advocated for a ban on open-weights models.” And: “All sufficiently capable models, open and closed, should go through mandatory safety testing.” (Anthropic, 27 Jul 2026). It did not sign NVIDIA’s open-weights letter; by 28 July “Anthropic and xAI are still not among” its 150-plus signatories (NYU Shanghai RITS; NVIDIA letter (PDF)).

Funding & Backers

• Series A (2021), $124M: “led by Jaan Tallinn, technology investor and co-founder of Skype. The round included participation from James McClave, Dustin Moskovitz, the Center for Emerging Risk Research, Eric Schmidt, and others.” (Anthropic). See Jaan Tallinn and Eric Schmidt.
• Series B (April 2022), $580M: Sam Bankman-Fried led the round, “joined by Caroline Ellison, Jim McClave, Nishad Singh, Jaan Tallinn and the Center for Emerging Risk Research” (TechCrunch, 29 Apr 2022).
• The overlap between Anthropic’s early money and the philanthropic network that funds AI-safety institutions (Moskovitz, Tallinn, the Center for Emerging Risk Research) is mapped on the research page and in the Holden Karnofsky file.

Political spending.

• $40 million to Public First Action: $20 million in February 2026 and $20 million more on 21 July 2026, “made exclusively to support Public First Action’s public education and policy mission,” which Anthropic says “cannot be used to influence the election of any candidate” (Anthropic, 21 Jul 2026).
• The CEO’s personal gift: FEC records show Dario Amodei gave $1,000,000 on 4 May 2026 to “Public First” (committee C00930503), a super PAC registered with the FEC as independent-expenditure-only (FEC receipts, Public First; FEC committee page). That super PAC is a separate committee from Public First Action Inc., the organization the company funded.
• Lobbying: Anthropic’s in-house lobbying-disclosure reports show $1,560,000 in Q1 2026 and $1,970,000 in Q2 2026 (LDA, Anthropic Q1 2026; LDA, Anthropic Q2 2026), with outside firms reporting separately, among them Ballard Partners at $380,000 for Q2 (LDA, Ballard Partners for Anthropic, Q2 2026).

Actions & Leadership Choices

The pacing call. On 12 September 2026 Dario Amodei published “We Must Pace the Frontier.” It argues: “The most effective method of pacing is via regulation that targets all US frontier AI companies, as that covers even those who are unwilling to cooperate voluntarily.” It asks the government to “issue a narrow waiver for certain kinds of safety conversations” on antitrust grounds, offers embedded external evaluators “Desks in our offices, access badges, and company laptops,” with the right to publish “without editorial control by Anthropic,” and insists that “We should approach any global pacing decision, especially in the near term, in such a way that protects the lead of the US and its allies.” (darioamodei.com).

At the Security Council. Amodei briefed the UN Security Council remotely on 23 September 2026: “At Anthropic, we believe we have the best track record in the industry for preventing these risks”; “We will slow down as much as necessary in order to make sure that every successive AI technology that we release is actually safe.” He proposed a ban on using AI to make biological weapons, evaluation and verification systems “so that states can have visibility into frontier model capability and can verify each other’s commitments,” and “a notification system for AI incidents that are significant to global security” (UN Transcripts, S/PV.10228, machine-generated).

Signed the voluntary version. On 29 September Amodei signed the “White House Accord on Super Intelligence,” four layers of internal controls, an internal team, an independent external auditor and a board committee, voluntary by its own terms (NPR / AP).

Blacklisted, and both rulings. The Defense Department designated Anthropic a supply-chain risk in March 2026 after contract talks collapsed: “The DOD wanted Anthropic to grant the military unfettered access to its models across all lawful purposes, while Anthropic wanted assurance that its technology would not be used for fully autonomous weapons or domestic mass surveillance” (CNBC, 25 Sep 2026). Secretary Hegseth accused the company of attempting “to seize veto power over the operational decisions of the United States military” (same). A national-security lawyer’s reading at the time: the statutory authority “is not a sanctions authority,” and “the designation of an American owned and operated company appears to be without precedent” (Just Security, Tess Bridgeman, 2 Mar 2026). The Pentagon relied on two designations, litigated in two courts:

• N.D. Cal., 27 August 2026: Judge Rita Lin found the labeling was “unlawful retaliation” in violation of the First Amendment and “arbitrary and capricious,” and wrote that the government’s “words and deeds confirm that the challenged actions were based on a desire to make a public example out of Anthropic for its ‘arrogance’ in criticizing the government” and that “The empty invocation of national security is not a blank check to punish and retaliate against government critics.” TechCrunch reports she also noted the government continued “collaborating with the company’s new model, Mythos, for cybersecurity” (TechCrunch, 28 Aug 2026).
• D.C. Circuit, 25 September 2026, 2-1: Judge Gregory Katsas, joined by Judge Neomi Rao, upheld the other designation: “In our Republic, it is the President and the Secretary of War who must determine how best to balance the competing risks.” Judge Karen LeCraft Henderson dissented. The panel delayed its decision from taking effect so Anthropic can seek rehearing (CNBC, 25 Sep 2026).

Named by the President and his advisers. On Truth Social on 14 September 2026 the President wrote that the administration “has stopped AI ‘people’ from doing bad, or potentially bad, ’things,’ like Dario (Anthropic!), who is now pretending to be a ‘perfect little angel’” (Trump’s Truth archive of the post; quoted by CNBC). In October 2025 White House AI adviser David Sacks accused the company of “a sophisticated regulatory capture strategy based on fear-mongering” and called it “principally responsible for the state regulatory frenzy that is damaging the startup ecosystem”; co-founder Jack Clark called the attack “perplexing” and said “simple rules with thresholds protecting startups” could benefit “the entire ecosystem” (The Decoder).

Under investigation, and sued. On 30 September 2026 a senior FTC official confirmed a “broad probe into the safety of AI systems, including Anthropic and OpenAI,” into “allegations of unfair or deceptive acts” (ABC News). Anthropic PBC is the lead-named defendant in Buist v. Anthropic, PBC, No. 3:26-cv-10693 (N.D. Cal., filed 18 September 2026), the consumer class action alleging that the labs’ public agreement to slow down is itself anticompetitive; answer due 15 October (CourtListener docket; PBS / AP).

Leadership choices. The CEO writes the policy essays, briefs the Security Council, signs the accord, and personally funds a super PAC; a co-founder (Jack Clark) answers the White House’s critics. A former Open Philanthropy chief executive works at the company (Holden Karnofsky). The company’s government relations are a CEO-led argument made in public, not a back-channel, and the government’s answer has also come in public: in the courts, on Truth Social and from the FTC.

CONDUCT: CONTESTED — ASKS FOR THE RULES, BLACKLISTED BY THE BUYER. Anthropic asks governments for the power to slow or block frontier deployments, funds the campaign for it, and says the US lead over China must be protected while it happens. It has kept that position while the Pentagon blacklisted it, and it has won one of the two court rulings on that blacklist and lost the other. Whether the brake is a safety mechanism or a moat is the argument the record does not settle. Both readings cite the same documents.

Sources: Anthropic — $124M Series A, 28 May 2021; TechCrunch — $580M Series B, 29 Apr 2022; Anthropic — Donating another $20 million to Public First Action, 21 Jul 2026; Anthropic — Our position on open-weights models, 27 Jul 2026; Dario Amodei — We Must Pace the Frontier; UN Transcripts — S/PV.10228 (machine-generated); NPR / AP, 30 Sep 2026; CNBC — appeals court upholds designation, 25 Sep 2026; TechCrunch — first court win, 28 Aug 2026; Just Security — Tess Bridgeman, 2 Mar 2026; Trump’s Truth — post of 14 Sep 2026; The Decoder — Sacks accuses Anthropic; ABC News — FTC probe; CourtListener — Buist v. Anthropic, PBC; PBS / AP — slowdown lawsuit; FEC — Public First receipts ≥$1M; FEC — Public First committee; LDA — Anthropic Q1 2026; LDA — Anthropic Q2 2026; LDA — Ballard Partners for Anthropic, Q2 2026; NYU Shanghai RITS — open-weights letter; NVIDIA — Open Weights and American AI Leadership (PDF); AI Magazine — benchmark figures.