Author: Ben Thompson Publication: Stratechery Date: March 2, 2026 URL: https://stratechery.com/2026/anthropic-and-alignment/
"Anthropic and Alignment" is a March 2, 2026 Stratechery essay by Ben Thompson arguing that Anthropic's standoff with the Department of War over autonomous weapons and mass surveillance is, in Thompson's words, "intolerable and misaligned with reality." Thompson contends that if Dario Amodei is correct that AI is on a trajectory to equal or exceed nuclear weapons in geopolitical leverage, then a private US company seeking to dictate military terms to the elected US government is structurally untenable. The essay extends Thompson's earlier chip-strategy writing and accompanies the supply-chain-risk designation episode involving Defense Secretary Hegseth.
Central argument
Thompson frames international law as ultimately a function of power, with nuclear weapons fundamentally tilting that balance. On that premise, he argues that if Amodei is right about AI reaching nuclear-weapon-equivalent leverage, the United States faces a binary choice. Either Anthropic accepts subservience, with Congress and the President deciding military use; or the US government destroys Anthropic or removes Amodei. Thompson states that he does not want the second outcome but argues the logic is unavoidable.
The standoff
Thompson recounts the factual record underlying the dispute. The Trump administration designated Anthropic a supply-chain risk and ordered all federal contractors to cease commercial activity with the company; per the essay, the designation had not previously been applied to an American company. Separately, OpenAI reached an agreement with the Department of Defense to use OpenAI models in classified settings, a status that until then only Anthropic had held.
Anthropic's stated red lines were mass domestic surveillance and fully autonomous weapons. The Department of War demanded that these safeguards be removed under an "any lawful use" clause and threatened to invoke the Defense Production Act to force compliance. Semafor reported that Anthropic raised concerns over Claude's appearance on screens during the Maduro raid, conducted via Palantir and AWS Top Secret infrastructure; Anthropic denied objecting to the operation.
Subsidiary arguments
Thompson advances three further positions in the essay.
On chip controls, he argues that Amodei's framing that "selling AI chips to China is like selling nuclear weapons to North Korea" (stated at Davos 2026) is more dangerous than Amodei realizes. Thompson reasons that if China cannot obtain AI, it becomes game-theoretically rational for China to destroy TSMC, which would make chip controls increase the risk of a Taiwan invasion rather than deter it.
On open source, Thompson argues that Anthropic's anti-open-source stance is short-sighted, offering three reasons that more AI is safer than less AI: closed AI vests unimaginable power in a small number of stewards; safety-by-stewardship ignores the structural incentives for proliferation; and the best defense against AI is more AI.
On Anthropic's conduct, Thompson characterizes the company as singularly focused on being "the one winner with all the power," with limited consideration of how others react.
Palmer Luckey contrast
Thompson endorses Palmer Luckey's "do you believe in democracy?" argument as a complementary frame, quoting: "Should our military be regulated by our elected leaders, or corporate executives?... It is easy to say 'But they will have cutouts to operate with autonomous systems for defensive use'... but you immediately get into the same issues and more — what is autonomous? What is defensive?"
Reception and relation to other sources
The essay's position runs counter to Anthropic's own statement of position, which maintained its red lines, and to the resulting Anthropic v. United States (Pentagon ban challenge) suit. Thompson's safety-via-proliferation reasoning aligns with the Open-Source AI / Open-Weight Models argument. The essay also sits opposite the The AI Grand Bargain — Ben Buchanan and Tantum Collins (Foreign Affairs, October 2025) position: Buchanan and Collins argue the state should coordinate with industry, whereas Thompson argues the state should dominate private holdouts. The piece explicitly references and extends AI Promise and Chip Precariousness — Ben Thompson (Stratechery, February 2025) (February 2025), Thompson's earlier chip-strategy essay.
Relationships
- depends-on: Ben Thompson
- related: Dario Amodei — primary subject
- related: Anthropic
- related: Palmer Luckey — endorsed counterpoint
- related: Anthropic v. United States (Pentagon ban challenge)
- related: AI and National Security
- related: AI Promise and Chip Precariousness — Ben Thompson (Stratechery, February 2025)
- contradicts: Anthropic official red-line position on autonomous weapons + mass surveillance
- supports: Open-Source AI / Open-Weight Models