Sen. Elizabeth Warren is poking around the Pentagon’s AI contracts, and some of the biggest names in tech are in her crosshairs. Nvidia, Oracle, Google, Microsoft, and AWS are all part of a sweeping set of agreements designed to bring artificial intelligence into classified military networks.

Warren formally requested detailed information on July 7 about the contracts, zeroing in on concerns about surveillance capabilities and autonomous weapons development.

What the Pentagon built

The Department of Defense confirmed agreements on May 1 with eight companies: Nvidia, Oracle, Microsoft, AWS, Google, OpenAI, SpaceX, and Reflection AI. The goal is integrating AI capabilities on classified networks categorized as Impact Level 6 and 7, which handle some of the military’s most sensitive data.

The Pentagon has framed the initiative around building an “AI-first fighting force” aimed at what it calls decision superiority.

Anthropic, the AI safety company that built Claude, was excluded from the agreements. The reason: Anthropic reportedly refused to comply with terms related to surveillance and autonomous weapons systems.

No specific financial terms or contract values have been disclosed.

Warren’s concerns and the transparency gap

Warren’s letters demand answers on several fronts. She wants to know what guardrails exist around AI-powered surveillance tools and what role these technologies might play in developing autonomous weapons systems.

This inquiry builds on the Pentagon’s prior partnerships with companies like OpenAI and SpaceX. Nvidia stands out as a particular focus due to what Warren’s office characterizes as the company’s flexible contract terms, suggesting Nvidia may have been especially accommodating to military requirements.