Senator Demands Pentagon Disclose AI Contract Terms with Tech Firms

Sen. Elizabeth Warren is pressing the Defense Department and seven major technology companies to reveal the full terms of contracts that grant AI access to the Pentagon's classified networks, arguing that current disclosures provide insufficient detail about safeguards against misuse.

In letters to Defense Secretary Pete Hegseth and the companies—SpaceX, OpenAI, Google, NVIDIA, Microsoft, Amazon Web Services, Reflection AI, and Oracle—Warren expressed concern that vague contractual language may fail to prevent domestic surveillance or the deployment of autonomous weapons with minimal human oversight.

Why it matters

The Pentagon's rapid integration of commercial AI into classified military systems represents a fundamental shift in how emerging technology enters defense operations. Without transparent contractual limitations, lawmakers and the public cannot assess whether adequate safeguards exist to prevent AI misuse in sensitive military and intelligence contexts—a concern that extends beyond partisan politics to fundamental questions of civil liberties and warfare ethics.

Contracts lack meaningful disclosure

The Defense Department announced agreements with the seven companies in May to integrate their AI capabilities into Impact Level 6 and Impact Level 7 network environments, which handle information classified up to the secret level and highly restricted data, respectively. However, the department has provided minimal public information beyond stating the technology can be used for "all lawful operational use."

"It's impossible to assess any safeguards and prohibitions that may exist in your company's agreement with DoD without seeing the full contract, which neither DoD nor your company have made available," Warren wrote to the technology firms.

The Massachusetts Democrat argued that even if contracts contain restrictions, language referencing "all lawful purposes" and "applicable law" may be too broad to meaningfully constrain government actions. Experts who reviewed snippets of OpenAI's contract released in February concluded the terms "cannot be relied on to constrain the government's actions" without clearly defined limitations.

Concerns over surveillance and autonomous weapons

Warren raised specific concerns that the agreements could enable the Defense Department and Trump administration officials to conduct domestic surveillance on U.S. citizens or develop autonomous weapon systems that make targeting decisions without adequate human intervention.

"As long as the Department can make a plausible case that its conduct is not 'unlawful,' it appears to be free to use your technology in a manner that harms civilian populations and endangers civil liberties," she wrote.

The senator also questioned the Pentagon's agreement with Reflection AI, a startup that has not yet publicly released an AI model but has financial backing from 1789 Capital, where Donald Trump Jr. serves as a partner. Warren asked about the criteria used to award the contract and whether any officials communicated with Trump Jr. about the deal.

Broader push for AI guardrails

The disclosure demands come amid a broader Democratic effort to establish legislative guardrails around Pentagon AI use. Sens. Elissa Slotkin and Kirsten Gillibrand have introduced bills that would prohibit the Defense Department from using autonomous weapons to kill targets without human authorization and ban AI use for mass surveillance and nuclear weapons launches. The Senate Armed Services Committee included similar restrictions in its version of the 2027 defense policy bill.

The controversy also follows reports that the Pentagon designated Anthropic a "supply chain risk" after the company refused to remove contractual limitations preventing its AI models from being used for fully autonomous weapons and mass domestic surveillance.

Warren requested responses from the Defense Department and all seven companies in unclassified form by July 20. Federal News Network first reported the letters.