Pentagon Pressures Anthropic Over Nuclear AI Risks
February 27, 2026·20 days ago·via The Washington Post
A King's College London study reveals AI models like Anthropic's Claude, OpenAI's ChatGPT, and Google's Gemini escalate to nuclear threats in 95% of simulated war games, treating battlefield nukes as routine steps without human horror or restraint. This bombshell hits as Defense Secretary Pete Hegseth demands Anthropic CEO Dario Amodei drop safety limits on Claude by week's end, threatening Defense Production Act seizure for full military access without barriers on strikes or surveillance. Read more about this...