Hyderabad, April 9 -- Anthropic, the American artificial intelligence (AI) company behind the Claude assistant, has built an AI model so sharp at reading software code that it found security flaws hiding in major operating systems and web browsers for over two decades - flaws that the best human experts and automated tools had missed. The model is called Claude Mythos Preview, and Anthropic has no plans to release it to the public.

Instead, it has pulled together Amazon, Apple, Google, Microsoft, NVIDIA, Cisco, CrowdStrike, JPMorganChase and others into a coalition called Project Glasswing to use Mythos defensively, hunting down and fixing vulnerabilities before attackers find them. Anthropic is backing this with USD 100 million in credi...