Mythos and Cybersecurity

Anthropic announced Claude Mythos Preview, a specialized AI model for discovering and weaponizing software vulnerabilities, restricting access to ~50 critical infrastructure vendors under Project Glasswing. The model found thousands of vulnerabilities across operating systems and browsers, weaponizing 181 Firefox exploits versus its predecessor's two. Security expert Bruce Schneier critiques the incomplete public disclosure—highlighting missing false-positive rates and the risk that restricted access may leave less-visible software (medical devices, industrial control systems) undefended while a skilled adversary could use Mythos as a force multiplier.