xAI Signs Pentagon Deal for Grok in Classified Systems — Anthropic Faces Ultimatum

The Deal

Elon Musk's AI company xAI has signed an agreement with the U.S. Department of Defense to allow military use of its Grok model in classified systems, a Defense official confirmed to Axios. The move marks a major shift in how advanced AI is deployed across the most sensitive areas of national security.

The Anthropic Standoff

Until now, Anthropic's Claude was the only AI model available in systems used for the military's most sensitive intelligence work, weapons development, and battlefield operations. The Pentagon had been pushing Anthropic to make Claude available for "all lawful purposes" — a standard the company refused to meet.

Anthropic has specifically insisted on blocking Claude's use for the mass surveillance of Americans and the development of fully autonomous weapons systems. Claude was previously used in operations such as the Maduro raid, through Anthropic's partnership with Palantir.

xAI's Terms

Unlike Anthropic, xAI agreed to the Pentagon's "all lawful use" standard. Grok now joins Google's Gemini and OpenAI's ChatGPT as models available in defense-accessible systems.

The Ultimatum

Defense Secretary Pete Hegseth is set to meet with Anthropic CEO Dario Amodei, with sources describing the meeting as an ultimatum. The Pentagon is reportedly considering branding Anthropic a "supply chain risk" — among other potential penalties — if the company refuses to lift all AI safeguards.

Broader Implications

This confrontation highlights the fundamental tension between AI safety principles and military requirements. How Anthropic responds could set a precedent for whether AI companies can maintain ethical constraints when facing pressure from defense clients.