---

The Department of Defense (DoD) recently issued an ultimatum to Anthropic, demanding unrestricted use of its AI systems for “all lawful purposes.” Crucially, the DoD has refused to rule out using these systems for mass domestic surveillance or fully autonomous lethal weapons.

In response to Anthropic prioritizing these safety redlines, the Department has threatened to retaliate by invoking the Defense Production Act or designating the company a “supply chain risk,” a label usually reserved for foreign adversaries.

We are gathering signatures from technology professionals, business leaders, scientists, civil society leaders, and concerned citizens to urge Congress to exercise oversight. Will you join us?

---

The Honorable Roger F. Wicker, Chairman
The Honorable Jack Reed, Ranking Member
Senate Committee on Armed Services

The Honorable Mike Rogers, Chairman
The Honorable Adam Smith, Ranking Member
House Committee on Armed Services

The Honorable Ted Cruz, Chairman
The Honorable Maria Cantwell, Ranking Member
Senate Committee on Commerce, Science, and Transportation

The Honorable Brian Babin, Chairman
The Honorable Zoe Lofgren, Ranking Member
House Committee on Science, Space, and Technology

Re: Congressional Oversight of the Department of Defense’s Use of Procurement Pressure to Remove AI Safety Guardrails

Dear Chairmen and Ranking Members:

We write as technology professionals, business leaders, scientists, civil society leaders, and concerned citizens to respectfully request that these Committees exercise their oversight authority in response to the Department of Defense’s recent ultimatum to Anthropic — and to the broader questions that ultimatum raises about executive branch authority over private companies and the constitutional limits on domestic surveillance.

The core fact that warrants your attention is this: the Department of Defense has refused to rule out using a commercial AI system for mass domestic surveillance of American citizens or in fully autonomous lethal weapons systems — and has threatened legal retaliation against a private company for asking the Department to do so.

Specifically, the Secretary of Defense demanded that Anthropic allow military use of Claude for “all lawful purposes” without any specific restrictions, and declined to assure Anthropic that its products would not be deployed for domestic mass surveillance or in fully autonomous lethal weapons systems operating without human authorization. When Anthropic declined, the Department threatened to invoke the Defense Production Act to compel compliance or to designate Anthropic a “supply chain risk” — a designation otherwise typically reserved for foreign adversaries. 

These developments raise two sets of questions that warrant congressional attention.

First, the proposed “all lawful purposes” contractual clause does not resolve the constitutional questions — it defers and obscures them. It does not define what uses are lawful, bind any particular official, or provide any outside party a means of enforcement. It commits only to what the executive branch itself determines, at any given moment, to constitute a lawful purpose. The actual constitutional questions — whether AI-assisted mass surveillance of American citizens comports with the First and Fourth Amendments; whether the delegation of lethal force to autonomous systems without human authorization is consistent with law and constitutional norms — remain entirely unaddressed by both Congress and the courts. The executive branch should not be permitted to resolve them unilaterally and informally through a catch-all procurement clause. Frontier AI systems capable of aggregating, connecting, and analyzing information at scale represent an expansion of surveillance capability unlike anything previously available. Whether and under what legal constraints such systems may be turned against American citizens are questions that belong before Congress — not questions to be papered over in a procurement contract.

Second, the Department’s threatened use of the Defense Production Act raises serious and unresolved separation of powers questions. This law was designed to direct the production of goods and materials in support of national defense. It has not previously been used to compel a company’s staff to build and maintain products against their own judgment or even their conscience, and Congress has not authorized this use. 

The threatened “supply chain risk” designation raises a separate concern. That authority was established to protect the defense industrial base from foreign adversaries and has been applied to companies such as Huawei on that basis. Applying it to a U.S. company for abiding by its own safety commitments — in the absence of any objective, statutory risk determination — would represent a significant expansion of that authority’s intended scope, with broad implications for many other American enterprises that rely on that company’s technology.

That the Administration is simultaneously asserting — via its supply chain threat — that Anthropic’s product is so dangerous that it cannot safely be integrated into the U.S. government’s systems and — via its DPA invocation — that its product is so essential that it must be forced into the U.S. government’s systems exposes the insincerity of each asserted claim and reveals an executive branch seeking to manufacture leverage to control a private company’s design choices, ethical standards, and expression. That this action could be highly damaging to a leading private sector company and what is right now America’s most important economic industry only heightens the stakes.

We do not write in opposition to the use of AI in support of legitimate national security objectives. We write because the questions raised by this episode — about the scope of executive procurement authority, the constitutional limits on domestic AI surveillance, and the appropriate role of Congress in defining those limits — are consequential and unresolved. We respectfully urge these Committees to examine these questions through hearings and, where appropriate, legislation.

Respectfully submitted,