Sen. Elizabeth Warren, a Massachusetts Democrat, accused the Pentagon of what she called a potentially retaliatory move after the Defense Department formally blacklisted AI company Anthropic as a supply chain risk. In a joint statement with Sen. Andy Kim, a New Jersey Democrat, Warren framed the designation as an abuse of government power tied directly to Anthropic’s refusal to strip safety restrictions from its AI models for military use. The confrontation has quickly escalated into a legal battle, with Anthropic filing suit to block the blacklisting, and it raises hard questions about whether federal procurement tools designed for national security are being turned into instruments of coercion against private technology firms.
Warren and Kim Call the Blacklist “Extortion”
The two senators did not mince words. In their joint statement, Warren and Kim said reports that Defense Secretary Pete Hegseth may invoke the Defense Production Act to force an American company to remove guardrails on AI models amounted to government extortion. Their argument is straightforward: Anthropic declined to let the military use its AI technology without safety restrictions, and the Pentagon responded by cutting the company off from all federal contracts.
That sequence, the senators contend, is not a legitimate exercise of supply chain risk authority. It is punishment. Warren described the government’s actions as an attempt to coerce Anthropic into abandoning the very safeguards the company built to prevent misuse of its AI systems. The statement specifically tied the blacklisting to Hegseth’s earlier demands, suggesting a direct line from the ultimatum to the designation and warning that such tactics could chill responsible safety practices across the industry.
Hegseth’s Ultimatum and the Friday Blacklisting
The dispute did not begin with the formal designation. According to a person familiar with the matter, Hegseth warned Anthropic to let the military use the company’s AI technology as it sees fit. He set a deadline: comply or face loss of contracts and other consequences. Anthropic did not comply.
On Friday, President Trump ordered U.S. agencies to stop using Anthropic technology, and the Pentagon formally notified the company that it had been designated a risk to the defense supply chain, effective immediately. The Pentagon stated the designation applied for “all lawful purposes,” a broad phrase that could extend the ban well beyond current contracts to future bids and subcontracting arrangements. The speed of the escalation, from private warning to public blacklisting within days, is what gives Warren’s retaliation theory its force. When a company refuses a demand and then immediately faces a punitive designation, the chronology itself becomes evidence of motive.
Officials have defended the move in general terms as necessary to ensure military access to critical technologies without what they characterize as arbitrary constraints. But they have not publicly detailed any traditional security vulnerabilities associated with Anthropic’s products, such as compromised infrastructure or foreign control, leaving critics to infer that the core dispute is over use restrictions rather than espionage or sabotage risks.
The Legal Basis and Its Limits
The Pentagon invoked 10 U.S.C. Section 3252, a statute titled “Defense: supply chain risk.” The law was designed to let the Defense Department exclude vendors whose products or services pose genuine risks to military supply chains, such as foreign-manufactured components with potential backdoors or firms with ties to adversarial governments. It allows senior officials to act quickly, often using classified information, and to bar agencies from contracting with designated entities.
Anthropic is an American AI company. Its “risk” to the supply chain, as framed by the Pentagon, appears to stem not from compromised hardware or foreign entanglements but from the company’s decision to maintain restrictions on how its AI models can be used. That distinction matters. The statute gives the Pentagon broad authority to act on supply chain vulnerabilities, but critics argue that applying it to a domestic firm’s voluntary safety policies stretches the law far beyond its original intent. If the government can label any vendor a supply chain risk simply because it refuses to remove product limitations, the designation becomes a tool of commercial pressure rather than national security.
Legal experts note that Congress crafted Section 3252 in an era of concern about foreign-made electronics and software embedded in sensitive systems. Using it to compel changes in product design or content moderation for AI systems is largely untested. That novelty feeds uncertainty for both the government and industry: if courts uphold this expansive interpretation, the Pentagon will have a powerful new lever; if they reject it, they may sharply curtail the department’s discretion in future cases.
Anthropic Files Suit
Anthropic did not wait long to respond. The company filed a lawsuit to block the Pentagon’s blacklisting. In its complaint, Anthropic stated: “These actions are unprecedented and unlawful. The Constitution does not allow the government to wield its enormous” power in this manner. The truncated public quote suggests the company is making a First Amendment or due process argument, positioning the case as a test of whether the federal government can punish a private firm for exercising control over its own products and setting ethical limits on their deployment.
The lawsuit adds a judicial dimension to what had been a political and administrative fight. If a court grants an injunction, it would be the first time a federal judge has blocked a supply chain risk designation on grounds that it was retaliatory rather than security-driven. That precedent could reshape how the Defense Department interacts with the entire technology sector, forcing more transparency around the rationale for designations and potentially inviting greater congressional oversight.
Anthropic’s complaint also underscores the economic stakes. Being labeled a supply chain risk does not just cut a company off from the Pentagon; it can taint its reputation with other agencies and private customers who worry about regulatory exposure. For a fast-growing AI firm, exclusion from federal markets could chill investment, slow hiring, and push talent and partnerships toward competitors perceived as more aligned with government demands.
Bipartisan Concern Over the Precedent
Warren and Kim are not the only lawmakers paying attention. Senators from both parties have weighed in on the broader implications. Sen. Tim Scott, a South Carolina Republican, along with Sens. Mike Crapo, Mike Rounds, and Thom Tillis, have raised concerns about the risks this approach poses to innovation and civil liberties. Their involvement signals that discomfort with the Pentagon’s actions is not confined to one party or ideological camp, even as members differ on how aggressively to confront the Defense Department.
The bipartisan anxiety reflects a practical worry: if the government can blacklist any AI company that maintains safety restrictions the military finds inconvenient, other firms may preemptively weaken their own guardrails to avoid similar treatment. That dynamic could erode industry-led safety standards at the very moment policymakers are urging companies to take greater responsibility for the downstream impacts of powerful AI systems.
Some lawmakers also see a separation-of-powers issue. Congress has been debating how to regulate advanced AI, including whether to mandate certain safeguards for high-risk uses. Allowing the executive branch to use procurement law to pressure companies into loosening restrictions, critics argue, effectively lets the Pentagon make de facto AI policy in secret, sidestepping open legislative negotiations.
AI Safety, Military Needs, and the Road Ahead
The clash over Anthropic sits at the intersection of AI safety, national security, and corporate autonomy. The company has marketed its flagship model, Claude, as a system designed with strong guardrails to reduce harmful outputs. According to reporting on the dispute, Pentagon officials view some of those limitations as obstacles to developing and testing military applications, while Anthropic insists that loosening constraints could enable misuse well beyond any specific defense project.
For the Defense Department, access to cutting-edge AI is increasingly seen as essential to maintaining strategic advantage. For AI labs, however, the reputational and ethical costs of building tools that can be easily adapted for autonomous weapons, disinformation, or mass surveillance are mounting. The Anthropic case forces both sides to confront whether there is a stable middle ground in which companies can serve government customers without surrendering their own safety principles.
As the lawsuit proceeds and congressional scrutiny intensifies, other AI firms will be watching closely. A court ruling in favor of the Pentagon could embolden agencies to demand more pliable systems, while a victory for Anthropic might encourage companies to adopt firmer red lines around military and intelligence use. Either outcome will reverberate far beyond one blacklisted vendor, shaping how the United States balances the imperatives of security, innovation, and the responsible governance of artificial intelligence.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
