Warren Opens Investigation into Pentagon’s Designation of Anthropic As National Security Risk, New OpenAI Contract

Warren Opens Investigation into Pentagon's Designation of Anthropic As National Security Risk, New OpenAI Contract

Warren is concerned that "DoD is trying to strong-arm American companies into providing the Department with the tools to spy on American citizens and deploy fully autonomous weapons without adequate safeguards."

Text of Letters (PDF)

Washington, D.C. - U.S. Senator Elizabeth Warren (D-Mass.), a member of the Senate Armed Services Committee, opened an investigation into the Department of Defense (DoD)'s decision to designate Anthropic a "supply chain risk" after the company refused DoD demands to remove certain contractual guardrails preventing the misuse of its AI tools. The senator is also probing the contract that OpenAI announced with DoD just hours after Secretary Hegseth declared Anthropic a supply chain risk, since this contract appears to lack the necessary safeguards to protect Americans from government surveillance and prevent civilian harm in the course of military operations.

DoD is increasingly deploying AI for military use in both classified and non-classified systems. Until recently, Anthropic's Claude tool was reportedly "in daily use across most parts of the military." In late February, however, after Anthropic refused DoD's demands to eliminate two narrow contractual limitations on DoD's use of Anthropic's AI model Claude-one to prevent mass domestic surveillance and one to prevent the use of AI to deploy fully autonomous weapon systems-DoD designated Anthropic a "supply chain risk" to national security, an unprecedented action previously used to address risks from foreign adversaries such as China's Huawei.

In a letter to Secretary Hegseth, Senator Warren pushed him on "what appears to be retaliation by the Department of Defense against artificial intelligence contractors that seek contractual guardrails to prevent the misuse of their AI tools."

"DoD did not have to take such extreme actions: it could have chosen to terminate its contract with Anthropic or continued using its technology in unclassified systems," wrote Senator Warren. "Instead, it appears that you went beyond this approach and retaliated against the company by weaponizing longstanding statutes intended to protect against genuine national security threats."

Senator Warren also raised concerns that "DoD is trying to strong-arm American companies into providing the Department with the tools to spy on American citizens and deploy fully autonomous weapons without adequate safeguards." She pointed out that DoD's deal with OpenAI "does not appear to include the same safeguards that Anthropic sought" and may "allow DoD to use OpenAI's models in a manner that harms civilian populations and endangers civil liberties."

Senator Warren is pushing Secretary of Defense Pete Hegseth for answers on his decision to designate Anthropic as a "supply chain risk," his negotiations with Anthropic and OpenAI, and the extent to which AI is used in our military for fully autonomous weapons and mass domestic surveillance.

The same day DoD announced its designation of Anthropic, OpenAI and DoD announced that they had negotiated an agreement for DoD to use OpenAI's technology in classified networks. While OpenAI has claimed that its contract contains safeguards to prevent the same redlines set by Anthropic, legal experts have expressed concern that OpenAI's contractual language contains numerous generalities and carve-outs that could allow DoD to use OpenAI's tools for both domestic surveillance and fully autonomous weapons.

In a letter to Sam Altman, CEO of OpenAI, Senator Warren probed the company on its recently announced contract with the Department of Defense.

"I am concerned [that] you appear to have rushed into an agreement with Secretary Hegseth that gives him and other Trump Administration officials free rein to engage in domestic surveillance - including spying on U.S. citizens exercising their legal rights - or build autonomous weapon systems that have enormous power to make decisions about targeting without human intervention," wrote Senator Warren.

In fact, Altman admitted that the haste with which OpenAI announced its agreement with DoD "looked opportunistic and sloppy." Additionally, the portions of the OpenAI-DoD contract that have been made public raise significant concerns that DoD could use OpenAI's technology to impinge on Americans' civil liberties or employ autonomous weapons that could harm civilians.

"Meaningful constraints to prevent civilian harm in warfare are a moral, legal, and strategic necessity, but nothing in this agreement stops DoD from using OpenAI's tools to employ AI-enabled weapons of war in a manner that threatens civilians," Senator Warren wrote. Senator Warren also warned that the agreement may allow OpenAI's technology to be used to undermine "Americans' democracy, freedom, and privacy."

Senator Warren is pushing for answers from both Secretary Hegseth and Sam Altman by April 6, 2026.

###