A journalist outlet called The Intercept sued the government to get documents, and after that lawsuit, they got over 400 pages showing exactly how deep OpenAI, Anthropic, Google, and xAI are working with the Pentagon.
Back in July 2025, all four companies signed contracts worth up to $200 million each. The paperwork lays out what they actually agreed to: building tools meant to sharpen military decision-making, advising the Pentagon on AI strategy, training military personnel, sitting in on war-game exercises, and even forecasting the risks of their own technology.
One detail stands out. Anthropic reportedly refused to sign off on unrestricted deployment across classified networks unless its own rules against autonomous weapons and domestic surveillance stayed intact. The documents suggest that pushback is part of why the Pentagon’s AI chief labeled Anthropic a supply chain risk back in February.
I want to be fair here and look at both sides. Militaries have always chased the most capable tools available, that part isn’t new or scandalous by itself. Every major power on earth is racing to put AI into its defense systems whether these four companies sign up or not.
What’s actually interesting is that we can see the paperwork at all. For once we get an honest look at where a company’s public promises line up with what it agreed to on paper, and where they don’t. These are the same companies that talk publicly about safety guardrails and responsible deployment, so seeing the actual contract language is worth more than any blog post.
I’d rather have companies making hard calls in the light than in the dark. A lawsuit that forces 400 pages into daylight is democracy doing its unglamorous, boring job, and that’s a good thing to see working.
——
Follow: @Ali Demi
Book your free AI clarity call, NOW!
https://buff.ly/TpWy277
——
Sources:
https://theintercept.com/2026/09/08/military-ai-weapons-contracts-openai-anthropic-google/
https://www.ibtimes.co.uk/ai-giants-pentagon-military-contracts-1818594
https://www.metatalks.ai/pentagon-contracts-ai-labs-forecast-own-risks/
Repost this. Thanks.