Anthropic's AI Model Is a Hacker's Dream

Anthropic's new AI model, Claude Mythos Preview, was built for general use but proved unexpectedly skilled at finding security vulnerabilities in virtually every major operating system. Fearing misuse, Anthropic restricted access to a select group of major companies like Google, Nvidia, and JP Morgan Chase through a program called Project Glasswing. These organizations use Mythos to identify and patch security gaps, then report findings back to Anthropic. The company warns that competitors, including OpenAI, are developing similar tools, and expects comparable technology to be publicly available within 12 months.