Musk's Grok goes classified with the Pentagon

The Pentagon just drew a new battle line in the AI industry: accept military use without guardrails, or get replaced.

Defense officials confirmed to Axios on Monday that Elon Musk's xAI has signed a deal allowing its Grok model to deploy in classified military systems. The New York Times first broke the news of the agreement.

Until now, Anthropic's Claude was the only AI system cleared for the Pentagon's classified environments, which handle ultra-sensitive intelligence, battlefield operations, and weapons development.

The "all lawful purposes" ultimatum

The shakeup comes down to a fundamental clash over safeguards. The Pentagon demands that frontier models be made available for "all lawful purposes."

Anthropic refused. The lab drew a hard line against allowing Claude to be used for mass surveillance of Americans or the development of fully autonomous weapons.

That pushback triggered a severe backlash. Defense Secretary Pete Hegseth is hosting Anthropic CEO Dario Amodei on Tuesday for what sources describe as a tense meeting. Hegseth plans to issue an ultimatum: drop the safeguards, or the Pentagon may label Anthropic a "supply chain risk" and impose harsh penalties.

xAI, meanwhile, had no problem agreeing to the military's "all lawful use" standard.

A messy transition ahead

While Grok is officially cleared to enter classified networks, defense officials admit that actually ripping and replacing Claude will be an uphill technical battle. Claude has already been woven into classified defense operations via partnerships with Palantir, including during the Maduro raid.

At the same time, the Pentagon is trying to bring other vendors into classified spaces. Grok, Google's Gemini, and OpenAI's ChatGPT are already running in unclassified networks.

The New York Times reported that Google is close to a classified deal for Gemini while OpenAI is not. A defense official pushed back on that framing, saying talks with both are active and the military expects both to sign eventually.

Still, sources noted that the Pentagon recently reached out to OpenAI with fresh urgency to reignite talks, though complex disagreements remain. The requirement to accept "all lawful purposes" remains non-negotiable for the military.

Why it matters

This is a defining moment for defense tech and Silicon Valley ethics. Musk's xAI has positioned itself as the pragmatic, no-questions-asked alternative to traditional frontier labs. For enterprise and government clients frustrated by safety filters, Grok is rapidly becoming the path of least resistance.

For the broader industry, the Pentagon's threat to designate Anthropic a supply chain risk sets a chilling precedent. Historically, tech giants held significant leverage when setting ethical boundaries for military contracts. By threatening blacklists, the Defense Department is signaling that ideological guardrails will no longer be tolerated in federal procurement.

We think this pivot carries serious operational risk for the military. Claude earned its spot inside sensitive operations like the Maduro raid because of proven reasoning and deep integration with Palantir. Grok may check the legal and compliance boxes Hegseth wants, but swapping a battle-tested model for a newer system purely to bypass ethical safeguards looks like a gamble on technical stability.