The Pentagon has granted xAI’s Grok access to classified defense networks, making it one of the first commercial AI systems cleared for classified government use. Senator Elizabeth Warren sent a formal letter to the Department of Defense pressing for justification of the decision, TechCrunch reported on March 16.

According to DoD statements cited in the report, Grok has been “onboarded” to classified infrastructure but is “not yet being used” in a classified setting. OpenAI received parallel access through a separate agreement. The dual clearance means the Pentagon now has two commercial AI providers with access to its most sensitive networks.

The Musk Conflict

Warren’s letter zeroes in on the conflict of interest created by Elon Musk’s simultaneous roles as xAI’s owner and a senior figure in the Department of Government Efficiency (DOGE). Musk’s DOGE position gives him visibility into government operations and spending decisions. His xAI company is now a cleared vendor for the same government.

The question Warren is asking is procedural but consequential: did the Pentagon follow standard procurement and security review processes when granting xAI classified network access, or did Musk’s proximity to the administration influence the decision?

”Onboarded” vs. Deployed

The Pentagon’s phrasing deserves attention. “Onboarded but not yet being used” suggests that Grok has passed security clearance and integration requirements but hasn’t been activated for classified workloads. In defense procurement, this distinction matters: clearance is the hard part. Once a system is approved for classified environments, deployment is an operational decision, not a security one.

That means Grok could go live on classified networks without any additional review or congressional notification. The clearance is the gate, and Grok is already through it.

Broader Context

This story broke during GTC 2026 week and was largely overshadowed by NVIDIA’s NemoClaw announcement and the OpenClaw acquisition coverage. But it represents a significant expansion of commercial AI’s role in national security infrastructure.

The Pentagon has historically relied on purpose-built defense AI from contractors like Palantir and Anduril. Clearing a consumer-facing AI model for classified use is a different category of decision, one that assumes commercial frontier models can meet the security, reliability, and auditability requirements of classified operations.

NCT has been tracking the Anthropic-Pentagon standoff, where the DoD designated Anthropic a “supply chain risk” after the company refused to remove safety restrictions for military use. The xAI clearance adds another dimension: the Pentagon is simultaneously pushing out the AI company that insists on safety constraints and pulling in one owned by a figure with an active government role.

Warren’s letter has not yet received a public response from the DoD.

Source: TechCrunch, March 16, 2026