Experts Say OpenAI Pentagon Contract Has Surveillance Loopholes

Former Pentagon Officials Challenge OpenAI's Contract Safeguards

An investigation published by The Intercept on March 8, 2026, has cast serious doubt on the protections OpenAI claims to have built into its contract with the Department of War. Multiple national security experts and former Pentagon officials say the contract language contains significant loopholes that could allow domestic surveillance and weaken meaningful restrictions on autonomous weapons use.

OpenAI secured the Pentagon contract in late February 2026, after rival Anthropic's deal collapsed over the company's refusal to enable mass surveillance and autonomous weapons capabilities. OpenAI CEO Sam Altman publicly stated that "two of our most important safety principles are prohibitions on domestic mass surveillance" and that the company had added stricter language on these issues. But the company has refused to release the actual contract text.

Legal Experts Identify 'Escape Hatch' Language

The core concern centers on the contract's use of qualifying terms like "intentionally" and "deliberately" before prohibitions on surveillance activities, and the phrase "consistent with applicable laws" as a baseline standard. Legal experts told The Intercept that these terms function as escape hatches that could render the safeguards meaningless in practice.

Brad Carson, who served as undersecretary of the Army during the Obama administration, was blunt in his assessment. Carson stated he was "not confident in the language at all" and told reporters he had "reluctantly come to the conclusion that this provision doesn't really exist, and they are just trying to fake it."

Carson explained that the word "surveillance" as used in the contract likely does not cover the activities most people would find concerning — such as using a large language model to build intelligence dossiers on private citizens using data pulled from federal and commercial databases.

"They're trying to blind you with complicated legal terms that ordinary people think mean something different entirely. But the lawyers know what it means. And the lawyers know that this is no guardrail at all." — Brad Carson, former undersecretary of the Army

The Transparency Problem

Alan Rozenshtein, a former Department of Justice attorney and current law professor, pointed out the fundamental issue: "There is nothing OpenAI can do to clarify this except release the contract." Without the actual text being public, all assurances from OpenAI executives remain unverifiable claims.

The investigation also highlighted the Pentagon's documented history of conducting activities that would commonly be understood as domestic surveillance — including purchasing Americans' location and browsing data without warrants through commercial data brokers. This directly contradicts claims by OpenAI chief strategy officer Jason Kwon that the Pentagon has no legal authority for such activities.

A former Pentagon AI official who reviewed the available contract language told The Intercept that the "consistent with applicable laws" clause is "the get out of jail free card right there," because the government routinely claims legal authority for surveillance programs that operate in gray areas of the law.

Altman Previously Acknowledged the Deal Was Rushed

The scrutiny comes after Sam Altman himself acknowledged problems with how the deal was announced. During a public Q&A session on X (formerly Twitter), Altman admitted the contract was done quickly "in an attempt to de-escalate the situation" following Anthropic's exit, and conceded it "just looked opportunistic and sloppy."

Altman also made a notable statement about the limits of corporate decision-making on matters of national security, saying: "I think you should be terrified of a private company deciding on what is and isn't ethical in the most important areas." He acknowledged that OpenAI executives "are not elected" and lack democratic accountability for decisions with national security implications.

What This Means for Tech Workers and Job Seekers

The controversy has already triggered real consequences for OpenAI. Caitlin Kalinowski, the company's head of robotics, resigned over concerns about the Pentagon deal's implications for surveillance and autonomous weapons. Nearly 900 tech workers from Google, OpenAI, IBM, and other companies signed an open letter urging the Pentagon to reverse Anthropic's supply-chain risk designation — a signal that the defense-AI debate is reshaping workplace culture across the industry.

For engineers and tech professionals evaluating employers, the OpenAI-Pentagon saga underscores a growing reality: a company's relationship with government defense agencies is becoming a material factor in talent recruitment and retention, alongside traditional considerations like compensation and technical challenge.