AI model finds security flaws nobody was looking for — including in systems you use daily
Anthropic's new AI model, called Mythos, has discovered previously unknown vulnerabilities in every major operating system and web browser, according to company officials. U.S. intelligence agencies are now evaluating the technology. This isn't about AI making existing security work faster — it's about AI finding entire categories of flaws that traditional methods miss.
Bottom Line
An AI model that can find security flaws by reasoning about how code works, rather than matching known patterns, has reportedly discovered vulnerabilities in all major operating systems and browsers. U.S. spy agencies are evaluating it. This represents a potential shift in how software weaknesses are discovered — and suggests some vulnerability classes may have been systematically invisible to traditional methods. The immediate question: does discovery outpace patching, or does this actually strengthen security by surfacing flaws that would otherwise stay hidden?