On August 31, 2026, OpenAI’s ChatGPT Mil and xAI’s Grok for Government went live on GenAI.mil, the Pentagon’s day-to-day generative AI portal — joining Google’s Gemini for Government as the platform’s third and fourth models. Per DefenseScoop and TechCrunch, the expansion reaches up to 3 million Department of Defense civilian and military personnel. One name is missing from that list, and it isn’t an oversight.

Anthropic’s Claude isn’t on GenAI.mil. That’s not new — this site covered Anthropic’s exclusion from the Pentagon’s classified networks back in May. What’s new is where the exclusion now shows up. GenAI.mil isn’t a classified system. It’s the unclassified portal DoD staff actually use for email drafts, policy memos, and logistics planning. The same fight that kept Claude off IL6 and IL7 top-secret networks has now kept it off the everyday tool, too.

Quick Summary: What Happened

Detail	Info
Date	August 31, 2026
What changed	ChatGPT Mil and Grok for Government joined Google’s Gemini on GenAI.mil
Reach	Up to 3 million DoD civilian and military personnel; 1.7 million+ unique users already onboarded
Security tier	Impact Level 5 (Controlled Unclassified Information)
Missing vendor	Anthropic’s Claude — absent from GenAI.mil entirely
Why	Same contract dispute behind Anthropic’s exclusion from classified IL6/IL7 networks: refusal to drop guardrails on mass surveillance and autonomous weapons
Sources	DefenseScoop · TechCrunch · NOTUS

Bottom line: The Pentagon’s Anthropic ban was never a classified-network technicality. It’s now the policy for every tier of DoD AI access, from top-secret down to the unclassified portal 1.7 million people already use.

What Is GenAI.mil?

GenAI.mil is the Department of Defense’s centralized, CAC-gated portal for accessing commercial frontier AI models on unclassified government data. It launched in December 2025 with Google’s Gemini as its only model, cleared to Impact Level 5 — the DoD security tier for Controlled Unclassified Information. As of August 31, 2026, it serves as the single front door for ChatGPT, Grok, and Gemini access across the department.

What Actually Happened

The rollout itself is straightforward. ChatGPT Mil is built for “document-heavy unclassified work,” per the Pentagon’s own description cited by Military Times — planning, policy drafting, logistics, and administration, scaled for the department’s full 3-million-person workforce. Grok for Government ships with what the Pentagon calls “adaptive reasoning modes,” customizable workspaces, and reusable “playbooks,” aimed at acquisition research and supply-chain management, according to NOTUS’s reporting.

Both new models cleared IL5 certification, the same tier Gemini has held since GenAI.mil’s December launch. The platform had already pulled in more than 1.7 million unique users out of its 3-million-person target base before ChatGPT and Grok arrived — proof this isn’t a pilot program nobody uses. It’s infrastructure DoD has actually adopted at scale.

What models are now available on GenAI.mil?

1. Google Gemini for Government — the platform’s original model, live since December 2025
2. OpenAI’s ChatGPT Mil — added August 31, 2026, built for document-heavy planning, policy, and administrative work
3. xAI’s Grok for Government — added the same day, positioned for acquisition research, supply-chain management, and multi-mode reasoning tasks

Anthropic’s Claude is not on that list, and it’s the only major U.S. frontier lab that isn’t.

The Same Fight, One Network Lower

Here’s the part that makes this more than a routine feature rollout.

Four days before ChatGPT and Grok landed on GenAI.mil, U.S. District Judge Rita Lin ruled that the Pentagon’s “supply chain risk” designation against Anthropic was illegal — “unlawful retaliation” against the company for enforcing its own usage policy, arbitrary and capricious, and a denial of due process. CNBC’s coverage quoted Lin writing that “the empty invocation of national security is not a blank check to punish and retaliate against government critics.” It was Anthropic’s first real court win in a fight that’s dragged on since February.

Four days later, the Pentagon expanded its enterprise AI portal to three vendors and Anthropic still wasn’t one of them.

That sequencing matters. This isn’t a case of Claude simply not having cleared IL5 accreditation yet, or a procurement queue Anthropic hasn’t reached. The exclusion persists through an active court ruling that called the underlying justification unlawful. Whatever paperwork changes as Anthropic’s lawsuit proceeds, the practical access question — does Claude reach DoD’s 3 million personnel — hasn’t moved.

The dispute underneath both bans is identical. Anthropic’s usage policy bars Claude from mass surveillance and fully autonomous weapons applications. The Pentagon wanted that carve-out removed for “all lawful” military use; Anthropic refused. That standoff kept Claude off IL6 and IL7 classified networks in May, when Defense Secretary Pete Hegseth labeled Anthropic a national security risk. The same standoff apparently applies at IL5, an unclassified tier built for routine memo-writing and logistics spreadsheets — work with no plausible mass-surveillance or autonomous-weapons dimension at all.

That’s the tell. If the guardrail dispute were really about specific high-risk use cases, an unclassified administrative tool would be the easiest place to make an exception. It wasn’t. The exclusion reads less like a targeted capability restriction and more like a blanket vendor decision that happens to track the same fight, regardless of what the actual GenAI.mil workload looks like.

Why This Matters

GenAI.mil isn’t a side project. It’s the tool that 1.7 million DoD employees already use for the unglamorous work — the memos, the acquisition research, the supply-chain spreadsheets that make up most of a bureaucracy’s actual output. Being excluded from that portal is a bigger commercial loss for Anthropic than being excluded from IL6/IL7, in pure volume terms. Classified networks are a narrower, higher-clearance population. GenAI.mil is everyone.

For federal AI vendors and the contractors who integrate around them, the signal is clear: the “all lawful use” standard the Pentagon set for classified access is now the standard at every tier, court rulings notwithstanding. OpenAI and xAI both accepted DoD’s terms and now have distribution to 3 million users at every classification level the department maintains. Google, already a $40 billion Anthropic investor, is running Gemini through the same unclassified pipeline — the same dual posture this site flagged in May, where Google ships its own model into Pentagon environments while its Claude stake sits outside them entirely.

For enterprise buyers watching from outside government, the pattern from May still holds and gets reinforced here: Anthropic’s refusal to bend its usage policy under pressure is a data point about how the company behaves, not just a Pentagon procurement footnote. A vendor whose guardrails survive contact with its single largest potential government customer, at every network tier that customer operates, is telling you something real about how it’ll behave on your account too.

The Bigger Picture

Three frontier labs — OpenAI, Google, and xAI — now have standing, department-wide distribution into the Pentagon’s daily workflow. That’s a meaningfully different competitive position than “cleared for classified contracts.” Classified access is prestige and a ceiling-of-trust story. GenAI.mil access is recurring usage, feedback data, and the kind of institutional habit-forming that’s hard to dislodge once 3 million people have an account.

Grok’s inclusion is the more interesting story here, separate from Anthropic. Six months ago, xAI was a distant fourth in enterprise credibility conversations. Now it’s shipping alongside OpenAI and Google inside the Pentagon’s primary AI portal, with GSA pricing reportedly as low as 42 cents through March 2027 for federal agencies broadly. Whatever the state of xAI’s competitive position against Claude, ChatGPT, and Gemini in the consumer and enterprise markets, it’s now cleared for the same government distribution tier as the two labs that have spent years building federal relationships.

Anthropic’s absence, meanwhile, keeps compounding. Each new GenAI.mil vendor addition and each new classified-network agreement widens the usage gap between Claude and its three main rivals inside the single largest AI customer in the country. The Project Glasswing guardrails that define Anthropic’s red lines aren’t changing. Neither, it seems, is the Pentagon’s response to them.

Our Take

The Aug 27 ruling and the Aug 31 rollout landing four days apart isn’t a coincidence worth reading too much into — GenAI.mil’s ChatGPT and Grok additions were almost certainly in the pipeline well before Judge Lin issued her opinion. But the juxtaposition still tells you something. A federal judge called the Pentagon’s justification for excluding Anthropic unlawful, and the practical exclusion didn’t so much as pause. That’s a distinction enterprise buyers should sit with: legal rulings on how a designation was made don’t automatically restore commercial access. The label can lose in court while the vendor decision stands.

We think Anthropic’s position remains the more defensible one on the merits — a usage policy that bars mass surveillance and autonomous weapons is a narrower ask than “all lawful use,” and a company enforcing that line consistently, including against its biggest potential customer, is behaving exactly as a values-driven vendor should. But defensible and commercially costless aren’t the same thing. GenAI.mil’s 1.7 million existing users are a real, quantifiable market that OpenAI, Google, and xAI are now inside and Anthropic is not. That gap doesn’t close because a judge agreed the Pentagon’s paperwork was wrong.

For federal AI watchers, the next signal to track is whether Anthropic’s court win translates into any actual procurement movement — a GenAI.mil listing, a classified-network agreement, anything concrete — or whether it stays a legal victory with no commercial follow-through. Right now it’s the latter.

Frequently Asked Questions

What is GenAI.mil?

GenAI.mil is the Department of Defense’s centralized portal giving up to 3 million civilian and military personnel access to commercial frontier AI models for unclassified work. It launched in December 2025 with Google’s Gemini and is certified to Impact Level 5, the DoD security tier covering Controlled Unclassified Information.

Which AI models are on GenAI.mil as of September 2026?

Three models: Google’s Gemini for Government (live since December 2025), OpenAI’s ChatGPT Mil, and xAI’s Grok for Government (both added August 31, 2026). Anthropic’s Claude is not available on the platform.

Why isn’t Claude on GenAI.mil?

Anthropic refused to remove usage-policy guardrails barring Claude from mass surveillance and fully autonomous weapons applications when the Pentagon requested unrestricted “all lawful use” access. That same dispute led the Department of Defense to exclude Anthropic from classified IL6/IL7 networks in May 2026, and it apparently extends to the unclassified GenAI.mil tier as well.

Did the court ruling against the Pentagon change anything for Anthropic?

Not yet, practically speaking. On August 27, 2026, U.S. District Judge Rita Lin ruled that the Pentagon’s “supply chain risk” designation against Anthropic was unlawful. Four days later, GenAI.mil expanded to three other vendors without adding Claude. The ruling addressed how the designation was made, not whether Anthropic gets platform access.

How many people use GenAI.mil?

More than 1.7 million unique users had onboarded as of the August 31, 2026 expansion, out of a target population of roughly 3 million DoD civilian and military personnel.

What’s the difference between this and the Pentagon’s earlier Anthropic ban?

The May 2026 exclusion applied to classified DoD networks at Impact Levels 6 and 7. GenAI.mil operates at Impact Level 5, an unclassified tier for routine administrative and planning work. Anthropic is now absent from both the classified and unclassified sides of DoD’s AI infrastructure.

Does this affect Claude’s availability outside government?

No. Claude’s commercial availability through Anthropic’s API, AWS Bedrock, and Google’s Model Garden is unaffected. This exclusion is specific to Department of Defense platforms.

---

Last updated: September 1, 2026. Sources: DefenseScoop — Grok, ChatGPT added to GenAI.mil · TechCrunch — The Pentagon now has its own version of ChatGPT and Grok · NOTUS — Pentagon Broadens AI Tools to Include ChatGPT and Grok · Military Times — The military’s ChatGPT is now live via the Pentagon’s GenAI platform · TechCrunch — Anthropic gets its first court win over the Pentagon’s supply-chain-risk label · CNBC — Judge blocks Pentagon blacklist of Anthropic.

Related reading: Pentagon Bars Anthropic: What It Means for Enterprise AI · Project Glasswing and the Mythos Risk Question · Google’s $40B Anthropic Bet: What Changes for Claude · Anthropic vs OpenAI 2026 · Grok 4.20 Review