Table of Contents
The Pentagon is preparing to shake up the artificial intelligence tools running inside its most sensitive classified networks, opening the door to Elon Musk’s Grok model as tensions mount with current provider Anthropic.
According to Axios, Musk’s AI firm xAI has signed an agreement that would allow Grok to operate within secure Defense Department systems. If implemented, the move would mark a significant shift in how the military manages AI inside classified environments tied to intelligence analysis, weapons development and battlefield planning.
Until now, Anthropic’s Claude has held exclusive clearance for those top tier systems. That access made the company a central player in high stakes defense workflows, embedding its model deeply into U.S. military operations.
But that relationship has grown strained.
At the heart of the dispute is a policy clash over how the military can deploy advanced AI tools. Defense officials want flexibility to use AI for any legally permitted purpose, including surveillance and applications connected to autonomous weapons.
Anthropic, however, has refused to remove certain guardrails on Claude, even when the system is used within government networks. The company has maintained limits on how its model can be applied, citing safety considerations.
That resistance has reportedly frustrated Pentagon leaders, some of whom have warned that Anthropic’s position could jeopardize its role as a trusted supplier.
Replacing Claude would not be simple. The model is woven into classified systems that support intelligence gathering and strategic planning. It has reportedly been used in sensitive missions, including operations involving foreign heads of state, through Anthropic’s partnership with Palantir.
Still, the Defense Department appears willing to consider alternatives.
Under its agreement with the Pentagon, xAI has accepted terms allowing Grok to be used for any lawful government purpose. That clears a major hurdle for deployment inside classified networks.
Even so, some defense officials privately question whether Grok is ready to match Claude’s level of technical maturity in highly sensitive contexts. Grok is not yet widely viewed inside defense circles as battle tested at the same scale.
Grok already operates in unclassified Defense Department systems, alongside OpenAI’s ChatGPT and Google’s Gemini. Both companies have also discussed expanding their models into classified environments, and those conversations are said to be gaining urgency.
Last year, the White House approved Grok, ChatGPT, Gemini and Claude for government use. Claude was the only model authorized for the most sensitive classified systems.
xAI unveiled a government ready version of Grok in mid 2025, hoping to position it as a serious player in federal AI contracts. But almost immediately, it ran into trouble. In one widely shared exchange, the chatbot produced extremist and antisemitic remarks, triggering backlash and raising uncomfortable questions about how well the system was being monitored and controlled. The controversy unfolded shortly after a public disagreement between Musk and President Donald Trump over federal spending.
Now the Pentagon faces a complex decision. It must balance performance, operational control and ethical risk as AI becomes more deeply embedded in U.S. defense infrastructure.
Whether Grok ultimately replaces or simply complements Claude, the outcome will help shape the next chapter of military AI in the United States.
