Pentagon brands Anthropic national security risk, AI ties collapse fast
Anthropic’s Claude AI was the only system cleared on Pentagon classified networks via Palantir since 2024. A $200M contract supported intelligence assessment, with reports of use in the Iran war for target identification.
For months, the Pentagon demanded full AI access for any lawful use. CEO Dario Amodei resisted, seeking limits on autonomous weapons and mass surveillance. Anthropic agreed to cyber and missile defense, but safety guardrails remained disputed.
Just before 5 PM Friday, President Trump called Anthropic “Leftwing nut jobs,” threatening a federal ban. Defense Secretary Hegseth posted on X, accusing “arrogance and betrayal,” and moved toward a supply-chain risk designation.
Hours later, Sam Altman secured an OpenAI Pentagon deal for classified networks, replacing Anthropic. Elon Musk’s xAI also signed a classified agreement. A tech group including Nvidia and Apple urged restraint, but rivals filled the gap.
On Wednesday, the Pentagon labeled Anthropic a “supply-chain risk to national security,” previously used only for foreign adversaries. Effective immediately, Anthropic was banned from defense work, with a 6-month transition for existing services.
Amodei said Anthropic will challenge the ruling in court, calling it legally unsound. Experts warned the US treated it worse than China’s DeepSeek. Hudson Institute’s Sobolik criticized the move; analysts said it harmed security for “essentially no gain.”
