FOLLOW US

The chief executive of artificial intelligence company Anthropic has said the firm will continue defending its internal safety policies despite growing tensions with the United States Department of Defense over the military use of advanced AI systems.

In an exclusive interview with CBS News, Anthropic CEO Dario Amodei explained that the company remains open to working with the U.S. military but only if its core ethical and technical boundaries are respected.

“We are still interested in working with them as long as it is in line with our red lines,” Amodei said, referring to internal safety safeguards designed to limit certain applications of the company’s technology.

The disagreement emerged after Anthropic requested formal restrictions on how its AI model could be used in defense operations. The company has emphasized two primary concerns.

First, Anthropic wants to prevent large-scale surveillance of American citizens through government use of its AI systems. Second, the company seeks limitations on potential development or deployment of autonomous weapon systems that could operate without meaningful human oversight.

The Pentagon, however, has maintained that existing federal law already restricts mass surveillance of U.S. citizens and that internal military policies already limit fully autonomous combat weapons.

Defense officials argue that written contractual restrictions are unnecessary because operational safeguards already exist inside the military command structure.

The dispute intensified after federal authorities set a deadline for negotiations. When agreement was not reached, the administration directed federal agencies to discontinue use of Anthropic technology.

The decision followed a declaration that Anthropic would be considered a potential national security supply chain risk, restricting military contractors from continuing commercial engagement with the company.

Military technology officials stated that government users should rely on alternative AI platforms considered more aligned with defense requirements.

Emil Michael, the Pentagon’s chief technology officer, said the military had attempted compromise. “At some level, you have to trust your military to do the right thing,” he said, adding that security planning must also account for strategic competition in global AI development.

Anthropic leadership argues that AI development is advancing faster than legal and regulatory frameworks governing its use.

Amodei warned that emerging technology could enable new forms of data analysis or operational automation that society has not yet fully regulated.

Mass surveillance remains a primary concern. The company believes government agencies could potentially combine privately sourced data with advanced machine learning analytics.

On autonomous weaponry, Anthropic does not categorically oppose the technology but believes reliability and accountability challenges remain unresolved.

“We don’t want to sell something that we don’t think is reliable, and we don’t want to sell something that could get our own people killed or that could get innocent people killed,” Amodei said.

He added that responsibility for decisions made by fully autonomous weapons is still legally and morally ambiguous.

U.S. defense officials have argued that technological competition, particularly with rival global powers, requires rapid innovation.

Military representatives pointed to foreign AI development programs as justification for flexible operational policies.

Anthropic, however, has pushed for what it describes as narrow safety exceptions rather than broad unrestricted deployment.

The company says its models have not been shown to encounter the two high-risk areas it wants protected: large-scale surveillance of domestic populations and autonomous lethal targeting.

The dispute has also entered the political sphere, with some officials criticizing Anthropic’s position.

Defense Secretary Pete Hegseth described the company as “sanctimonious” and suggested its leadership was attempting to influence military policy through commercial technology agreements.

Mr. Trump publicly accused the company of ideological bias, saying, “Their selfishness is putting AMERICAN LIVES at risk, our Troops in danger, and our National Security in JEOPARDY.”

Anthropic rejected those characterizations and maintained that its objective is to protect both technological innovation and national interests.

Amodei said, “Disagreeing with the government is the most American thing in the world… We are patriots. In everything we have done here, we have stood up for the values of this country.”

If negotiations fail, the Pentagon is expected to gradually remove Anthropic technology from military systems over a six-month transition period.

Officials have described the planned shift as part of efforts to move toward alternative technology providers.

The company has indicated it may challenge any formal supply chain risk designation through legal channels if issued.

Amodei stated that Anthropic has not yet received official documentation of such designation but would respond through court action if necessary.

Amodei suggested that long-term governance of artificial intelligence should involve legislative oversight.

However, he acknowledged that legislative bodies often move slower than technological development.

“For right now, we are the ones who see this technology on the front line,” he said, emphasizing the company’s technical expertise.

The dispute reflects broader global debate about how artificial intelligence should be integrated into defense systems.

Experts say AI technology is becoming a strategic national resource comparable to energy or semiconductor infrastructure.

Supporters of stricter guardrails argue that ethical deployment of AI is necessary to maintain democratic accountability.

Critics warn that excessive restrictions could slow innovation and weaken national security capabilities.