Anthropic tries to repair Pentagon ties after AI used in Iran war

0
Anthropic tries to repair Pentagon ties after AI used in Iran war

Anthropic is trying to repair its relationship with the Pentagon after a dispute over how the military can use its artificial intelligence technology, according to a new report. At the same time, the U.S. military continues using the company’s AI model in operations tied to the war with Iran.

CBS News reports Anthropic CEO Dario Amodei told investors Tuesday the company is still in talks with the Defense Department after negotiations broke down over limits Anthropic wanted placed on its software.

“We’re still talking to them to try to de-escalate the situation and come to some agreement that works for us and works for them,” Amodei said during remarks at the Morgan Stanley Technology, Media and Telecom Conference in San Francisco. CBS obtained audio of the comments.

The talks follow a rupture between Anthropic and the Pentagon over the use of the company’s AI model, Claude. The dispute escalated after Anthropic pushed for written restrictions preventing the military from using the system for mass surveillance of Americans or for fully autonomous weapons.

FABRICE COFFRINI / AFP via Getty Images

President Donald Trump responded by directing federal agencies to stop using Anthropic’s technology, giving departments six months to phase it out. Defense Secretary Pete Hegseth also labeled the company a “supply chain risk,” a designation that restricts government contractors from working with the firm.

Anthropic disputes the designation and says it plans to challenge the move in court.

AI system remains embedded in military operations

Despite the phaseout order, the Pentagon continues relying on Claude inside a military artificial intelligence platform known as the Maven Smart System.

Two sources familiar with the military’s AI programs confirmed to CBS News that Claude was used during the U.S. strikes tied to the Iran conflict and remains part of the system. The Washington Post has also confirmed its use by the Pentagon.

U.S. Navy photo/Handout via REUTERS

Maven analyzes large volumes of intelligence from satellites, surveillance platforms and other classified sources. The system helps planners identify and prioritize potential targets and can generate strike options.

People familiar with the system told The Post it produced hundreds of potential targets during the opening phase of the Iran campaign.

Vantor/Handout via REUTERS

Military planners have integrated Maven across multiple commands over the past year. The AI model summarizes intelligence reporting, supports logistics planning and helps analyze battlefield information.

Dispute centers on limits for AI use

Anthropic says it sought explicit restrictions on how the government could deploy its technology. Company leaders argued certain uses would conflict with its policies and with what they describe as American values.

Amodei said Anthropic supports national defense but does not view itself as having an operational role in military decision-making.

“We’ve never questioned specific military operations,” he told investors. “We don’t see ourselves as having an operational role.”

Pentagon officials counter that federal law already prohibits mass surveillance of Americans and limits the use of fully autonomous weapons. During negotiations, the Defense Department offered written acknowledgments of those limits.

“At some level, you have to trust your military to do the right thing,” Pentagon chief technology officer Emil Michael told CBS News.

Pentagon reliance complicates phaseout

The Defense Department’s reliance on the system complicates the planned phaseout of Anthropic’s technology.

Officials familiar with the military’s AI programs say replacing Claude inside the Maven system could take months. The platform is already widely used across the armed forces, with thousands of personnel relying on it for intelligence analysis and operational planning.

The system’s speed is central to its use. AI tools process intelligence inputs and develop targeting options much faster than traditional planning methods.

That capability has already been deployed in multiple military operations. The Iran campaign marks the first time the technology has been used during a major active conflict.

Ella Rae Greene, Editor In Chief

Leave a Reply

Your email address will not be published. Required fields are marked *