The planned Trump-Amodei dinner demands public scrutiny

A court ruling did not end the dispute

On Friday, Sept. 25, the D.C. Circuit rejected Anthropic’s challenge to a Department of War action removing its Claude system from the department’s supply chain. The ruling was 2–1. The majority found the agency’s action within its authority under the relevant statute; Judge Karen LeCraft Henderson dissented, disputing the majority’s reading of the law’s definition of supply-chain risk, as the court’s opinion explains.

The opinion also describes a separate California court decision that set aside a different designation under a different law. The appeals court ruling therefore did not settle every legal dispute involving the company and the government.

The timing makes the proposed dinner worth scrutiny, but it does not show that the ruling or the dispute was discussed. The conflict centers on Anthropic’s limits on military use of its systems, including domestic mass surveillance and weapons that operate without human involvement.

Defense Secretary Pete Hegseth’s position, reported by Axios, is that the military should not let a company dictate the terms of operational decisions. That is a serious argument about who controls military operations. But it does not, on its own, settle which uses should be permitted or what safeguards should apply.

This is not a dispute between the government and a company without public ties. The Pentagon awarded Anthropic a $200 million contract in July 2025 to develop AI capabilities for national security, CBS News reported. When a private company supplies technology for military use, its limits and the government’s procurement decisions become matters of public concern, not just terms between a buyer and a vendor.

Those limits are not minor contract details. Domestic mass surveillance raises questions about government power over people at home; weapons operating without human involvement raise questions about who retains control over consequential decisions. The court’s ruling on statutory authority does not answer whether public safeguards are adequate.

The government and Anthropic both owe answers

The administration’s arguments should be distinguished, not blended together. Hegseth has argued for the Pentagon’s control over operational decisions; separately, a White House official told Axios that the administration’s goals include U.S. leadership in advanced AI, consumer protection, innovation and competitiveness.

President Trump has also argued that the United States should not slow AI development while competing with China, Euronews reported. The case for speed and military flexibility deserves an honest hearing. But competition and leadership do not by themselves explain what evidence supports a particular use, or where the administration would draw limits.

Anthropic’s position should not become a substitute for public oversight, either. CEO Dario Amodei has urged the AI industry to slow development so safety measures can catch up, the Associated Press reported. That concern deserves serious consideration, but a company’s warnings and its own safeguards should be examined rather than treated as the final word.

The public needs answers neither side can supply simply by invoking its own authority. What evidence supports the Pentagon’s security rationale? Which protections would govern military uses involving domestic surveillance or weapons without human involvement? Who decides whether safeguards are sufficient, and what oversight applies after a system is deployed?

The White House should explain how its stated aims for AI policy translate into rules for government use. Anthropic should explain how its safeguards work in practice and what limits it believes should bind the government. Neither company assurances nor a general claim of operational need should end that debate.

The public deserves a record

Axios also reported that Trump and House Speaker Mike Johnson planned to meet with leading AI executives on Tuesday, Sept. 29. That plan raises a straightforward question: what policy issues will officials discuss, and what will the public be told about the decisions that follow? The available reporting does not establish the meeting’s format or agenda.

We urge the White House to state whether the planned dinner with Amodei took place. If it did, the administration should release a readout identifying who attended and whether military AI, policy or Anthropic’s status came up. Sensitive operational details need not be made public for the basic subject of a policy discussion to be disclosed.

Anthropic should publish its own account of any policy discussion. That is a transparency measure, not a presumption of wrongdoing. When government decisions shape how powerful AI systems may be used, the public should not have to infer policy from private access.