Anthropic refuses Pentagon’s offer; US War Secretary says Amodei has “God-complex”

Anthropic, the AI company, has turned down the Pentagon’s newest proposal to change its contract, saying the suggested changes do not address the company’s concerns about AI being used for mass surveillance or fully autonomous weapons.

The disagreement centers on Anthropic’s AI system, Claude, which is the first AI approved to be used on the military’s classified network. The Pentagon wants fewer restrictions on how the AI can be used, while Anthropic wants to keep safeguards in place.

Defense Secretary Pete Hegseth reportedly told Anthropic CEO Dario Amodei that if the company doesn’t allow its AI to be used “for all lawful purposes,” the Pentagon would cancel its $200 million contract. He also warned that Anthropic could be labeled a “supply chain risk,” a serious classification usually reserved for companies linked to foreign adversaries.

View this post on Instagram

A post shared by HumNewsEnglish (@hum_news_english)

Anthropic said the Pentagon’s new offer was presented as a compromise, but included legal language that could let the safeguards be ignored.

Amodei said, in a detailed blog post yesterday, “I strongly believe AI can be crucial in defending the United States and other democracies against autocratic threats. But there are some cases where AI could harm, not protect, democratic values.” He said that technologies like mass surveillance and autonomous weapons are “beyond what AI can safely do today.”

Amodei added that these exceptions have not stopped the military from using Anthropic’s models in other ways. “The Pentagon’s threats do not change our position: we cannot in good conscience agree to their request,” he wrote.

The Pentagon responded through Emil Michael, Undersecretary for Research and Engineering, who criticized Amodei on X (formerly Twitter), calling him a “liar” with a “God-complex”, saying he is trying to control the military and putting national safety at risk. “The Department of Defense will always follow the law and will not bend to any one tech company,” Michael wrote.

After Amodei’s post, many Anthropic employees voiced support for the company. Trenton Bricken, part of the technical team, said, “Time and again, I’ve seen Anthropic stand by its values. This is a clear example.” Data science manager Gian Segato wrote, “It’s now obvious why Anthropic’s founding was so important, and how serious things could have been otherwise.”