How AI is getting better at finding security holes
Summary
AI models have recently become much better at finding software security flaws, and Anthropic announced Mythos Preview — a powerful new model that it says can identify high-severity vulnerabilities across major operating systems and browsers. Anthropic is restricting Mythos Preview access to about 50 organizations via Project Glasswing, citing high misuse risk, while allowing broader releases of related models. Open-source maintainers and projects like the Linux kernel and cURL report AI-generated reports have moved from noisy hallucinations to high-quality vulnerability findings, helping them find and fix more issues. Security experts warn the same capabilities can be weaponized by attackers once open-weight models catch up, and they note current models are often better at finding bugs than safely fixing them.
Classifications
AskAI Classifications
LabelsLinked Companies
