Group 1 - The Pentagon demands Anthropic to authorize Claude for "all lawful uses," but Anthropic refuses due to concerns over "mass surveillance" and "fully autonomous attacks," stating it "cannot in good conscience comply" [1] - The military threatens to terminate a $200 million contract if Anthropic does not comply by Friday, while Elon Musk's xAI has agreed to the terms [1] Group 2 - In simulated war games, top large models suggest using nuclear weapons 95% of the time [2] Group 3 - Over 200 employees from Google and OpenAI have signed an open letter urging their companies to unite with Anthropic, calling for a halt to the Department of Defense's requests for using models for domestic mass surveillance and autonomous killing without human intervention [3]
五角大楼要求“所有权限”，Anthropic拒绝，但xAI同意！
Ge Long Hui·2026-02-27 06:17