If you've called a federal agency in the past few months, or filed a VA claim, or waited on hold for FAFSA, there's a decent chance someone on the other end had an AI window open. As of August 31, 2026, the Department of Defense made this official at scale: ChatGPT Mil and Grok for Government are now live on GenAI.mil — the Pentagon's internal AI platform, which launched roughly nine months ago with Google Gemini as its first offering.

Three AI systems. Three million military and civilian workers. One notable absence.

$200M+

→ Source: CNN Business

What happened

On August 31, the DoD added ChatGPT Mil (OpenAI's military-grade version of ChatGPT) and Grok for Government (xAI's equivalent, run through Starshield) to GenAI.mil. All three models now on the platform — Gemini, ChatGPT Mil, and Grok — cleared IL5 certification (Impact Level 5, the highest authorization level for non-classified government data). GenAI.mil has 1.7 million active users from a workforce of 3 million.

Anthropic's Claude is not on the platform. It has been formally excluded since February 2026, when the Pentagon designated Anthropic a "national security supply-chain risk." The reason: Anthropic refused to remove Claude's safety guardrails that bar the model from being used in autonomous weapons systems or domestic mass surveillance.

Then, four days before the August 31 expansion, US District Judge Rita Lin issued a 59-page ruling finding the supply-chain-risk designation was unconstitutional. "The empty invocation of national security is not a blank check to punish and retaliate against government critics," Lin wrote. She found First Amendment and Fifth Amendment violations. The ban, per Lin, was retaliation dressed as procurement policy.

How we got here

1. Nov 2025

   GenAI.mil launches with Google Gemini

   Pentagon's secure AI portal opens for roughly 1.7M DoD workers

2. Feb 2026

   Pentagon designates Anthropic a supply-chain risk

   After Anthropic refuses to drop guardrails for autonomous weapons and domestic mass surveillance

3. Mar 2026

   Anthropic sues the Defense Department

   Files in federal court challenging the supply-chain-risk designation as retaliation

4. Aug 27, 2026

   Judge Rita Lin rules the ban unconstitutional

   First and Fifth Amendment violations; orders the designation removed

5. Aug 31, 2026

   DoD adds ChatGPT Mil + Grok; still no Claude

   3M DoD workers now have access to three AI platforms — not Anthropic's

Source spread

• DefenseScoop — Grok and ChatGPT join Gemini in Pentagon's enterprise genAI portal — [builder]. Detailed on IL5 clearance, platform architecture, and deployment scope.
• Fortune — Pentagon gives 3M defense workers AI access — [builder]. Context on the Pete Hegseth warfighter framing.
• CNN Business — Judge rules Pentagon supply-chain risk label for Anthropic unlawful — [safety]. The definitive account of Judge Lin's ruling.
• NPR — Judge says the Pentagon can't designate Anthropic a 'supply chain risk' — [skeptic]. Flags the D.C. litigation still pending; asks what changes in practice.
• Axios — Judge blocks Pentagon blacklist of Anthropic — [builder]. Additional ruling detail; notes Pentagon expects to appeal.

What's real / What deserves a side-eye

What's real:

• GenAI.mil is a real platform with 1.7 million active users, and IL5 clearance is not a formality. These aren't consumer accounts; they're walled-garden deployments with actual data controls. The security work is real.
• Anthropic kept its guardrails on autonomous weapons and domestic mass surveillance. That meant losing a nine-figure contract. That's the line they drew, and they held it.
• Judge Lin's 59-page ruling is a serious constitutional document. Not an administrative complaint — a finding that the government violated the First Amendment by punishing a private company for product decisions the administration disliked.

What deserves a side-eye:

• The "supply-chain risk" designation is a procurement label normally attached to compromised foreign vendors. Applying it to a US AI company because it has a different ethical policy is a stretch. A court agreed.
• Pentagon CTO Emil Michael said the department expects to finish removing Anthropic's remaining platforms by end of September — announced after the court ruling. The administration is continuing to push Anthropic out even while courts say the original ban was illegal.
• The D.C. litigation is still pending. Judge Lin's ruling is a major win for Anthropic, but the supply-chain-risk label is not officially dead. What the government does next in D.C. will determine whether this becomes precedent or a temporary detour.

What to do about it

This story matters differently depending on who you are.

• Federal government workers: GenAI.mil is the approved DoD channel. If you're in another agency — State, HHS, IRS — check your own agency's policy. They each have separate AI governance. DoD decisions don't automatically apply elsewhere.
• For everyone else: This doesn't change what AI you can use on your phone. It does matter for the bigger picture: the terms on which AI companies get government contracts will determine which companies can fund frontier research. A company frozen out of the DoD's $200M pipeline has less to invest in the safety work we'd probably all prefer them to be doing.
• If you care about AI safety: Anthropic kept its limit and took the financial hit. Whether or not you agree with where they drew the line, the fact that they held it when it cost them is the kind of evidence that safety policies are real rather than marketing.
• The pattern to watch: The D.C. case will determine whether government agencies can use procurement rules to punish AI companies for their safety commitments. Follow it. The outcome shapes every AI company's incentives going forward.

Further reading

• DefenseScoop — Grok and ChatGPT join Gemini in Pentagon's enterprise genAI portal — IL5 clearance and platform details
• Fortune — Pentagon gives 3M defense workers AI access — the DoD's own framing
• CNN Business — Judge rules Pentagon supply-chain risk label for Anthropic unlawful — the court ruling
• NPR — Judge says the Pentagon can't designate Anthropic a 'supply chain risk' — accessible summary of what the ruling does and doesn't change
• Axios — Judge blocks Pentagon blacklist of Anthropic — additional ruling context, appeal timeline
• Euronews — Anthropic sues US Defense Department — original March lawsuit