AI Ethics Clash: Anthropic Rejects Pentagon Demand For Mass Surveillance

A wide shot of a tense meeting in a dimly lit Pentagon briefing room. Pentagon officials in suits and uniform sit on one side of a long table, opposite Anthropic representatives with laptops glowing teal. A large screen displays a contract with "ANY LAWFUL USE" highlighted in red. The Department of the War plaque and a folder with a Venezuelan flag sticker are visible, highlighting the dispute over AI use in military operations.

Tensions grew for months between Anthropic and the Pentagon over how Claude could be used. The DoD demanded “any lawful use,” while Anthropic refused terms allowing mass surveillance and autonomous weapons.

US Defense Secretary Pete Hegseth stands behind his desk in a wood-paneled Pentagon office, pointing authoritatively at a document. The document is labeled "DEFENSE PRODUCTION ACT" with a red tab. A framed portrait of President Donald Trump is visible on the wall behind him, as he threatens to invoke the act against AI company Anthropic.

Pete Hegseth demanded a meeting with Dario Amodei, threatening to use the Defense Production Act and label Anthropic a “supply chain risk.” Spokesman Sean Parnell called media concerns “fake narratives.”

Over-the-shoulder shot of Pete Hegseth's dark silhouette facing Dario Amodei across a conference table. Amodei, the CEO of Anthropic, sits calmly with his hands clasped, his expression resolute. A contract lies on the table between them with "MASS SURVEILLANCE" and "AUTONOMOUS WEAPONS" circled in red, representing the red lines in the AI safeguards standoff.

On Tuesday, Hegseth met Amodei at the Pentagon. He threatened to remove Anthropic from the DoD supply chain. Amodei refused 2 red lines: using Claude for mass surveillance and autonomous weapons.

A close-up of a laptop screen and hands reviewing a contract late at night. The screen shows a legal document with "COMPROMISE LANGUAGE" highlighted in yellow, but critical text is highlighted in red with phrases like "may be disregarded." A hand holds a red pen, paused over a printed copy, revealing the Pentagon's deceptive contract wording.

On Wednesday night, the DoD sent updated contract terms. Anthropic said it showed “virtually no progress,” as safeguards could still be ignored. A former DoD official told the BBC the legal case was “extremely flimsy.”

Dario Amodei sits in his bright, modern San Francisco office, looking directly at the viewer with a calm, principled expression. He holds his right hand up in a "stop" gesture. On his desk, printed cards read "MASS DOMESTIC SURVEILLANCE" and "FULLY AUTONOMOUS WEAPONS," each crossed out with a bold red X, symbolizing his refusal to compromise on AI safety.

On Thursday, Dario Amodei said Anthropic would rather lose the Pentagon contract than allow mass surveillance or autonomous weapons. He added current AI systems are not reliable enough for fully autonomous weapons.

A wide shot of Dario Amodei walking away from the shadowy, imposing Pentagon building toward a bright, golden horizon. He glances back, carrying a laptop bag. Pete Hegseth watches from the Pentagon entrance with arms crossed. Broken chains labeled "SURVEILLANCE" and "AUTONOMOUS WEAPONS" lie on the path, symbolizing Anthropic's choice to uphold democratic values over a defense contract.

Amodei offered to help the Pentagon shift to another AI provider and suggested joint R&D on reliability, but the DoD declined. He said Anthropic would not provide products that risk warfighters or civilians.