NSA Runs Anthropic's Mythos AI Despite Pentagon's Supply-Chain Risk Label

The National Security Agency is running an artificial intelligence system built by Anthropic, the AI safety company, even as the Pentagon has flagged the startup's supply chain as a potential national security risk—a contradiction that raises questions about whether American intelligence agencies have adequate oversight of their own AI deployments. The arrangement, which has not been previously reported in detail, suggests that classification decisions and procurement rules may not be keeping pace with the speed and complexity of AI adoption across the federal government. Current and former officials familiar with the matter say the NSA's use of Mythos, Anthropic's large language model, highlights a broader vulnerability: the government lacks clear protocols for vetting and monitoring the AI systems it relies on, even as it races

Yellow.com·Sep 12