Anthropic’s Mythos model found vulnerabilities in classified US government systems, official says - The Washington Post

An AI security tool developed by Anthropic found vulnerabilities in classified U.S. government systems. This highlights both the value and risks of AI in critical national security infrastructure.