Anthropic Refuses Pentagon Demand to Remove AI Security and Safety Guardrails
Artificial intelligence (AI) firm Anthropic held firm to its founding principles this week, refusing to remove safety and security guardrails for its services in the face of mounting pressure from the U.S. Department of Defense (DOD).
DOD Secretary Pete Hegseth had demanded that Anthropic, in its contracts with DOD, remove prohibitions on using its Claude product to create fully autonomous weaponry or domestic mass surveillance programs. If Anthropic did not comply by 5:01 p.m. on 27 February, Hegseth had threatened to discontinue DOD’s use of Anthropic and use national security powers to further penalize the company.
Anthropic CEO Dario Amodei announced in a statement Thursday evening that the company would not concede to Hegseth’s demand. Amodei wrote that Anthropic could not “in good conscience” grant DOD’s request and that in a narrow set of cases, AI can undermine rather than defend democratic values.
“Some uses are also simply outside the bounds of what today’s technology can safely and reliably do,” Amodei wrote.
Amodei explained that his company supports the use of AI for lawful foreign intelligence and counterintelligence missions but using it for mass domestic surveillance is incompatible with democratic values.
“AI-driven mass surveillance presents serious, novel risks to our fundamental liberties,” Amodie wrote. “To the extent that such surveillance is currently legal, this is only because the law has not yet caught up with the rapidly growing capabilities of AI.”
Amodei also added that partially autonomous weapons can be used to defend democracy—such as how they are being used in Ukraine in its ongoing war with Russia. He acknowledged that fully autonomous weapons in the future may be critical to U.S. national defense.
“But today, frontier AI systems are simply not reliable enough to power fully autonomous weapons,” Amodei explained. “We will not knowingly provide a product that puts America’s warfighters and civilians at risk.”
Amodei wrote that without proper oversight, “fully autonomous weapons cannot be relied upon to exercise the critical judgment that our highly trained, professional troops exhibit every day. They need to be deployed with proper guardrails, which don’t exist today.”
If Anthropic did not comply with the DOD’s request, Hegseth had threatened to designate the company as a “supply chain risk” and to invoke the U.S. Defense Production Act to force Anthropic to remove its safeguards.
“These latter two threats are inherently contradictory: one labels us a security risk; the other labels Claude as essential to national security,” Amodei added.
The DOD did not respond to Security Management’s request for comment on this article. On X, however, Sean Parnell, a DOD spokesperson, wrote Thursday evening that the department does not want to use AI to develop autonomous weapons that operate without human involvement.
“Here’s what we’re asking: Allow the Pentagon to use Anthropic’s model for all lawful purposes,” Parnell posted. “This is a simple, common-sense request that will prevent Anthropic from jeopardizing critical military operations and potentially putting our warfighters at risk. We will not let ANY company dictate the terms regarding how we make operational decisions.”
The phrase “all lawful purposes” is at the heart of the disagreement between DOD and Anthropic. The U.S. Congress has failed to pass legislation to govern how AI is used in a national security context. Instead, the DOD has simply put its own policy limits on the development and use of autonomous weapons. This means that DOD could simply change its policies to allow the development and use of fully autonomous weapons or to create mass surveillance systems.
“If Congress had legislated guidelines on autonomous weapons and surveillance, Anthropic would likely be far more comfortable selling its systems to the military—and the [Defense Production Act] threat would have never arisen,” wrote Alan Z. Rozenshtein, associate professor of law at the University of Minnesota Law School, in a piece for Lawfare. “The question of what values to embed in military AI is too important to be resolved by a Cold War-era production statute.”
The clash between DOD and Anthropic was initiated in January 2026 when Hegseth issued an AI strategy memorandum that directed all DOD AI contracts incorporate standard “any lawful use” language within 180 days—which contradicted Anthropic’s existing contract with the DOD.
The DOD awarded Anthropic a transaction agreement with a $200 million ceiling in July 2025, alongside other awards to OpenAI, Google, and xAI. The contract was designed to allow Anthropic to prototype frontier AI capabilities that advance U.S. national security.
Since then, the U.S. government has been using Anthropic’s models on classified networks, at National Laboratories, and to create custom models for national security customers. The DOD currently uses Claude for intelligence analysis, modeling and simulation, operational planning, and cyber operations.
If the DOD designates Anthropic as a supply-chain risk, it will initiate a time-consuming process of protective measures and likely take at least three months to remove the company’s presence from Pentagon systems.
“Operators would have to reconfigure data inputs that they are feeding into models, re-examine how to share data in real-time with the intelligence community which also uses Claude widely, and re-validate that replacement models were functioning as the military expected it to,” according to Defense One.
This is not the first time that Anthropic has made a business decision that also poses U.S. national security ramifications. In the past, the company cut off the Chinese Communist Party’s (CCP’s) use of Claude and shut down CCP-sponsored cyberattacks that attempted to abuse Claude.
