Pentagon labels Anthropic a national security threat over its AI ethics limits

The US Department of Defense has officially added Anthropic to its list of supply chain risks, citing concerns that the company could shut off its technology in the middle of military operations. The decision stems from Anthropic's own policy: a set of so-called "red lines" that spell out how and when the company may restrict or halt its AI systems.

At its heart, the dispute is about corporate control over critical technology. Like many other AI companies, Anthropic has drawn up a set of ethical principles that limit how its models can be used in dangerous contexts. Among them are conditions under which the company reserves the right to forcibly shut its systems down. That is exactly what alarmed the Pentagon.

For military agencies, any dependence on a technology supplier that can cut off service unilaterally is a vulnerability. In defense, critical systems have to keep running no matter where the vendor stands. If AI tools are built into intelligence, logistics, or combat chains, the developer's ability to switch them off creates a risk similar to what the military calls a single point of failure.

An official Defense Department spokesperson confirmed that the existence of the "red lines" was the key reason Anthropic received the risk designation. In effect, the department's position is this: a company whose ethical limits allow it to walk away from cooperation in combat conditions cannot be considered a reliable defense partner.

Anthropic is known as one of the most principled AI companies when it comes to deploying its models safely. Its Responsible Scaling Policy (RSP) sets it apart from many competitors. The company has publicly stated that it will not help build weapons of mass destruction and reserves the right to turn down contracts that conflict with these principles. Now those same principles have led to an official clash with a federal customer. The situation calls into question whether responsible AI companies can work with the defense sector at all without sacrificing their ethical commitments.

The broader context matters. The US armed forces are actively integrating AI into a wide range of systems, from intelligence analysis and logistics planning to tactical decision support. According to public sources, Anthropic already holds contracts with US government agencies, including through Amazon Web Services, which uses Claude models in its government cloud offerings. That makes the situation especially delicate: the company is both an active supplier and an officially designated risk.

The precedent matters for the whole industry, too. If the Defense Department consistently rejects vendors with ethical limits, AI companies will face pressure to abandon such principles in exchange for access to large government contracts. The alternative scenario is the emergence of two separate ecosystems: military AI systems without ethical limits and civilian ones with them.

For Anthropic's investors and partners, this is a signal that the company's ethical stance carries a real cost in money and market opportunities. For regulators and the public, it raises a pointed question: should AI companies have the right to refuse the government, and if so, under what conditions? The tension between a developer's ethical control and national security demands is only growing, and the Anthropic case will likely not be the last.