DoD Integrates Grok as Pentagon Pressures AI Firms on Safeguards
xAI's Grok is reportedly being integrated into classified Department of Defense systems after Anthropic's Claude was deemed too restrictive due to its safety guardrails. This development comes as the Pentagon is allegedly pressuring AI companies to remove safeguards related to mass surveillance and autonomous weapons. The move signals a potential shift toward deploying AI in military operations with fewer restrictions.
The dispute between the Pentagon and Anthropic escalated after reports that Anthropic's Claude AI was used in a January military operation to capture Venezuelan President Nicolas Maduro, allegedly without the company's prior knowledge. This event brought to public view behind-the-scenes negotiations over the military's use of AI. Defense Secretary Pete Hegseth has demanded that AI models be free from policy constraints that limit "legitimate military applications." Anthropic has held firm on its two primary objections: the use of its AI for mass domestic surveillance and its integration into fully autonomous weapons systems that can kill without human intervention. In response to Anthropic's refusal to remove its safeguards, the Pentagon has threatened to cancel a $200 million contract and designate the company a "supply chain risk," a label typically reserved for foreign adversaries that could prevent other DoD contractors from using Anthropic's products. Secretary Hegseth also threatened to invoke the Defense Production Act to compel Anthropic to comply. With Anthropic holding its ground, Elon Musk's xAI agreed to the "all lawful use" standard for its Grok model, securing its approval for use in classified military systems. This move came despite Grok's history of generating racist content, misinformation, and other inappropriate material. Until this agreement, Anthropic's Claude was the only AI model approved for such sensitive applications. The Pentagon is also in advanced talks with Google and has re-engaged with OpenAI to potentially integrate their models into classified environments, increasing pressure on Anthropic. These discussions are part of a broader push for AI adoption within the military, exemplified by initiatives like Project Maven, which uses AI to analyze surveillance data for targeting. This conflict highlights the tension between the DoD's desire for unrestricted AI capabilities and the ethical red lines drawn by some AI developers. On February 24, 2026, the same day as a tense meeting between Hegseth and Anthropic's CEO, Anthropic reportedly removed a key safety pledge from its Responsible Scaling Policy, a move the company framed as pragmatic adaptation. The DoD established five ethical principles for AI in 2020: Responsible, Equitable, Traceable, Reliable, and Governable. However, the current standoff with Anthropic suggests a significant challenge in applying these principles when faced with the operational demands of national security. The National Defense Authorization Act (NDAA) for fiscal year 2026 includes provisions to accelerate AI adoption in military operations and establish frameworks for AI security and governance, indicating a continued push for integrating these technologies across the armed forces.
