AI Finds 10,000+ Security Bugs in One Month

A powerful new AI tool is making the internet safer for everyone by finding hidden security flaws at a speed humans never could.

Anthropic's Claude Mythos Preview, an unreleased AI model, discovered more than 10,000 software vulnerabilities just one month after launching Project Glasswing in April. The cybersecurity initiative partners with major tech companies to find and fix security bugs before hackers can exploit them.

The results have been remarkable. Mozilla found and fixed 271 vulnerabilities in Firefox using Mythos, 10 times more than it discovered in an older browser version using a different AI model. Cloudflare identified 2,000 bugs, with 400 rated as high or critical severity. Microsoft found so many issues that it announced patch releases will "continue trending larger for some time."

Partners using Mythos Preview are finding bugs more than 10 times faster than their previous methods. Most have discovered hundreds of critical or high-severity vulnerabilities in their software, issues that could have left users vulnerable to cyberattacks.

When Anthropic tested Mythos on 1,000 open-source projects, the AI found 23,019 total vulnerabilities. Of those, 6,202 were rated high or critical severity, the kinds of flaws that pose serious security risks.

The technology is so powerful that Anthropic hasn't released it publicly yet. The company says no one, including itself, has developed strong enough safeguards to prevent the model from being misused by bad actors. They plan to release similar models once those protections exist.

The Ripple Effect

The impact extends far beyond individual companies. Amazon Web Services, Apple, Google, NVIDIA, JPMorganChase, and Palo Alto Networks are all using Mythos to strengthen their security. That means billions of people using these services will benefit from stronger protection against cyber threats.

Anthropic is working to expand Project Glasswing's availability to the US government and other nations. This partnership could help protect critical infrastructure, government systems, and national security from increasingly sophisticated cyberattacks.

The initiative comes at a crucial time when AI-powered attacks are becoming more advanced. Using AI to defend against AI-driven threats creates a new frontier in cybersecurity that could keep everyone safer online.

Finding bugs this quickly means fewer opportunities for hackers to breach systems and steal data, protecting everything from personal information to financial records to national security secrets.