Anthropic's Mythos AI: Public Release & Cybersecurity Impact

Anthropic has confirmed its Mythos-class AI models will be available to the public in the next few weeks. The company says it's making rapid progress on safety protections for these models. Mythos is a large language model with advanced cybersecurity capabilities. It can reportedly find thousands of software vulnerabilities overnight. This ability has raised cybersecurity concerns among executives and world leaders. Mythos Preview testing has already found thousands of serious software vulnerabilities. For example, Cloudflare found around 2,000 bugs in its systems using Mythos, with nearly 400 being high or critical severity. Cloudflare also noted Mythos Preview's false positive rate was better than human testers. Currently, Mythos is available only in preview through Project Glasswing, an initiative with major tech companies like Amazon Web Services, Apple, and Google. These partners have reportedly seen their bug-finding rate increase by more than ten times. This rollout means powerful AI for cybersecurity could soon be widely accessible.