Musk’s xAI Wins Pentagon Deal as Anthropic Faces AI Access Threat

Pentagon Secures Alternative AI Partner as Dispute with Anthropic Intensifies

The U.S. Department of Defense has reached an agreement with xAI, Elon Musk’s artificial intelligence company, to allow the military to utilize its Grok model within classified systems. This move comes as the Pentagon is locked in a contentious dispute with Anthropic over safeguards and potential restrictions on the use of its AI model, Claude, and is reportedly considering declaring Anthropic a “supply chain risk.”

Until now, Claude has been the sole AI model operating within the military’s most secure environments, supporting critical intelligence work, weapons development, and battlefield operations. However, Anthropic has resisted the Pentagon’s demand for “all lawful purposes” access, specifically objecting to the potential for its technology to be used for mass surveillance of American citizens and the creation of fully autonomous weapons systems.

xAI has agreed to the Pentagon’s terms, allowing for broader use of its Grok model, according to a Defense official who confirmed the agreement. The New York Times first reported the finalized deal. XAI did not respond to requests for comment.

The timing of xAI’s agreement is significant, as Defense Secretary Pete Hegseth is scheduled to meet with Anthropic CEO Dario Amodei on Tuesday. Sources indicate that Hegseth intends to present Amodei with an ultimatum: lift all safeguards or face potential penalties, including being designated a “supply chain risk.”

Defense officials acknowledge that transitioning away from Claude and integrating a new AI model will be a complex and challenging undertaking. Claude was recently utilized in a military operation – the capture of former Venezuelan President Nicolas Maduro – through a partnership between Anthropic and data firm Palantir.

The Pentagon is also actively engaging with other AI companies, including Google and OpenAI, to explore the possibility of expanding their access to classified networks. Google’s Gemini and OpenAI’s ChatGPT are currently available in unclassified military systems. Negotiations with both companies have been accelerated as the potential for a split with Anthropic looms.

While the New York Times reported that Google was “close” to a deal for classified use of Gemini and OpenAI was “not close,” a Defense official disputed this characterization. The official stated that discussions with both companies are ongoing and that the department believes both will ultimately agree to the “all lawful purposes” criteria. However, administration officials have emphasized that this standard is non-negotiable for both Google and OpenAI.

One source indicated that the Pentagon had “reached out to OpenAI to reignite talks with a new sense of urgency,” but cautioned that a deal remains elusive due to the complex issues involved. The Pentagon is hopeful that both companies will ultimately comply with its requirements, but acknowledges that further negotiations are necessary.

The situation highlights the growing tension between the Department of Defense and AI developers regarding the ethical and security implications of advanced artificial intelligence technologies. The Pentagon is seeking unfettered access to these tools for national security purposes, while some AI companies are hesitant to relinquish control over how their technologies are used, particularly in sensitive areas such as surveillance and autonomous weapons systems.