Statement from Dario Amodei on our discussions with the Department of War

Anthropic’s Stance and Immediate Reactions

• Many commenters praise Anthropic for refusing to support domestic mass surveillance and fully autonomous weapons targeting, seeing it as rare backbone in tech.
• Others note that Anthropic still proudly supports extensive military/intelligence use (planning, cyber, analysis), so this is a narrow objection, not anti‑war.
• Some are glad enough to subscribe or stay subscribed; others say this letter confirms they won’t use Anthropic because it is deeply integrated with the US security apparatus.

Moral Principle vs PR and Strategy

• A recurring split: one side views this as a genuine moral stand that risks huge revenue and “supply chain risk” designation; the other sees a savvy PR move and negotiation tactic.
• Critics highlight prior IP issues, doomer marketing, and recent loosening of Anthropic’s own safety policy to argue the company’s ethics are selective or opportunistic.
• Defenders counter that even if partially performative, refusing these two use cases under public threat still matters in practice.

“Department of War” Naming and Authoritarian Drift

• The use of “Department of War” triggers a long subthread: some say it accurately describes what the US military does and exposes euphemistic “Defense” framing.
• Others stress the name hasn’t been legally changed and see Anthropic’s adoption of the new label as appeasing an increasingly authoritarian administration.
• There’s debate over whether the US is already “fascist” or merely trending that way, with references to threats against companies, press, and dissenters.

Domestic vs Foreign Surveillance and Non‑US Users

• Non‑Americans are especially angry that Anthropic explicitly opposes domestic mass surveillance but explicitly “supports lawful foreign intelligence,” reading this as: privacy for US citizens only.
• Several point out long‑standing intelligence sharing (e.g., Five Eyes) makes “foreign vs domestic” a legal fiction: spying on foreigners often routes back to domestic surveillance.
• Some argue Anthropic is tailoring its argument to US constitutional law and domestic politics, not articulating a universal human‑rights position.

Autonomous Weapons and Military AI

• Many assume fully autonomous weapons are inevitable and note that landmines and some existing systems are already “autonomous” in practice.
• Others emphasize that once kill decisions are automated, democracy’s safeguard of a human military unwilling to fire on its own population erodes.
• Several note Anthropic’s framing: fully autonomous weapons “may prove critical” once reliable, suggesting today’s refusal is about current technical limits and liability, not a timeless ban.

Power, Contracts, and the Defense Production Act

• Commenters focus on the contradiction Anthropic flags: being simultaneously threatened as a “supply chain risk” and as essential enough to compel under the Defense Production Act.
• Some argue government can nationalize or requisition tech in wartime, making corporate “values” ultimately fragile; others think nationalization of an AI lab would trigger mass resignations and rapidly destroy its technical edge.
• A few stress this conflict exists only because Anthropic previously did choose to work with the Pentagon; refusing at all would have required a much earlier line in the sand.

Geopolitics, China, and Democracy Rhetoric

• The opening language about “defending democracies” and “defeating autocratic adversaries” draws fire: critics see it as US‑centric, Sinophobic, and blind to US‑backed abuses abroad.
• Others argue that deterrence against China, Russia, etc. requires top‑tier military AI and that refusing to help the US simply hands advantage to less constrained regimes.
• There’s no consensus: some prioritize constraining US power as the bigger threat to them personally; others prioritize maintaining US military/technological superiority.

Broader Anxiety About US Trajectory

• The thread widens into pessimism about US decline, erosion of democratic norms, and a “military‑industrial + surveillance” state that long predates this administration.
• Some see acts like Anthropic’s as one of the few encouraging signs of institutional resistance; others consider them cosmetic gestures within an unfixable system.