# OpenAI Pentagon Contract (2026)

**Type:** Military AI contract  
**Announced:** February 27, 2026  
**Parties:** OpenAI, U.S. Department of Defense  
**Status:** Active (amended March 2-3, 2026)  

## Overview

OpenAI's Pentagon contract, announced the same day Trump banned Anthropic from federal use, established the operative template for military AI contracts after governance-refusing alternatives were excluded. The contract accepts 'any lawful use' language that Anthropic refused, with voluntary red lines added as non-binding constraints.

## Timeline

- **2026-02-27** — OpenAI CEO Sam Altman announces Pentagon contract accepting 'any lawful use' language. Same day, Trump orders federal agencies to cease using Anthropic's AI technology and Secretary of Defense Pete Hegseth designates Anthropic a 'supply chain risk.'
- **2026-02-27 to 2026-03-02** — Public backlash. 1.5 million users quit ChatGPT over the deal. MIT Technology Review publishes 'OpenAI's compromise with the Pentagon is what Anthropic feared.'
- **2026-03-02** — Altman admits initial rollout appeared 'opportunistic and sloppy.'
- **2026-03-02 to 2026-03-03** — OpenAI amends contract to explicitly prohibit surveillance of 'U.S. persons' and ban 'commercially acquired' personal information. Critics note amendments still contain intelligence agency carve-outs.
- **2026-03-08** — The Intercept publishes 'On Surveillance and Autonomous Killings: You're Going to Have to Trust Us' characterizing OpenAI's approach as relying on voluntary trust rather than structural constraints.

## Contract Terms

**Core language:** 'Any lawful use' — permits deployment for any purpose authorized by existing statutes.

**Voluntary red lines (initial):**
- No mass domestic surveillance
- No directing autonomous weapons systems

**Amendments (March 2-3):**
- Explicit prohibition on surveillance of 'U.S. persons'
- Ban on 'commercially acquired' personal information
- Intelligence agency carve-outs preserved

## Critical Analysis

**EFF characterization:** The red lines are 'weasel words' because 'any lawful use' language permits broad data collection under current statutes.

**The Intercept framing:** 'You're Going to Have to Trust Us' — voluntary trust rather than structural constraints.

**MIT Technology Review:** 'OpenAI's compromise with the Pentagon is what Anthropic feared' — demonstrates that Anthropic's stand was not shared by the other major AI lab.

## Significance

The OpenAI deal establishes what 'military AI governance' looks like in practice when the governance-refusing option (Anthropic) is excluded: voluntary red lines, no constitutional protection, contractual rather than structural constraints, accepted surveillance loopholes. This is the baseline that future AI governance will be compared to.

## Sources

- NPR, 'OpenAI Announces Pentagon Deal After Trump Bans Anthropic,' February 27, 2026
- MIT Technology Review, March 2, 2026
- The Intercept, 'On Surveillance and Autonomous Killings: You're Going to Have to Trust Us,' March 8, 2026