Artificial intelligence lab Anthropic has made clear it has no intention of loosening its usage restrictions for military applications of its AI technology, a source familiar with the matter said, even as talks with the U.S. Pentagon continue following a high‑stakes meeting this week, reports Reuters.
The dispute came to a head on Tuesday when Anthropic CEO Dario Amodei met with U.S. Defense Secretary Pete Hegseth to address a months‑long disagreement over how Anthropic’s AI — particularly its flagship model Claude — can be used by the U.S. military.
Anthropic has stood its ground on safety constraints designed to bar the technology from being used for fully autonomous weapons targeting or domestic surveillance of U.S. citizens. The Pentagon, however, insists that its AI partners should allow all lawful military use cases without built‑in restrictions.
During the meeting, Hegseth delivered an ultimatum: accept broader access or face potential actions, including being designated a “supply‑chain risk” — a label typically reserved for foreign adversaries — or having the U.S. government invoke the Defense Production Act to compel changes to Anthropic’s usage policies. The company was reportedly given until Friday to respond.
Anthropic has maintained that its current safeguards do not impede the Defense Department’s present operations, and a spokesperson characterized Tuesday’s discussion as “continued good‑faith conversations” about balancing national security needs with responsible, ethical AI deployment.
The dispute highlights broader tensions as the Pentagon negotiates AI contracts with several large language model providers — including Google, OpenAI and xAI — that will shape future military use of AI across applications such as battlefield analysis, autonomous systems and cybersecurity. Anthropic was previously the only provider on classified networks, though the Pentagon recently reached an agreement with xAI to deploy its models across secure systems.
The standoff intensified after Pentagon officials grew concerned about Anthropic’s inquiries into how its AI was used in a military operation in Venezuela earlier this year, a source said. During the meeting, Amodei reportedly clarified that Anthropic had not raised objections with partners or the Pentagon about that event and reiterated the company’s ethical boundaries.
Legal experts say that if the Pentagon invokes emergency powers to override Anthropic’s policies, it could spark significant litigation, given the unprecedented nature of compelling an AI firm to change its usage rules under national security authority.
As negotiations continue, the clash exemplifies the evolving debate over how advanced AI technologies should be governed within national defence frameworks, balancing innovation, ethics and security imperatives.
Bd-pratidin English/ Jisan
