← Retour au blog
tech 22 May 2026

Project Glasswing: An Initial Update

Project Glasswing is revolutionizing software security by uncovering over ten thousand critical vulnerabilities with AI. Discover how this is reshaping the cybersecurity landscape.

Article inspired by the original source
Project Glasswing: An Initial Update ↗ www.anthropic.com

Introduction: Cybersecurity in the Age of AI

Cybersecurity is undergoing a transformation. With the rise of increasingly powerful artificial intelligence (AI) models, threats are evolving at a breakneck pace. It is in this context that Project Glasswing was launched, a collaborative effort to secure the world's critical software before these models can be misused.

Promising Initial Results

Since the project's inception, more than 50 partners have used the Claude Mythos Preview to uncover over ten thousand high- or critical-severity vulnerabilities in some of the world's most important software. Previously, progress in software security was limited by how quickly new vulnerabilities could be found. Today, the challenge lies in verifying, disclosing, and patching these vulnerabilities as swiftly as possible.

The Performance of Mythos Preview

The Mythos Preview tool has far exceeded experts' expectations. Although public disclosure of vulnerabilities follows the industry's 90-day convention, Glasswing partners have already seen an exponential increase in their bug-finding rates. For instance, Cloudflare discovered 2,000 bugs, 400 of which are high or critical severity.

Impact on Cybersecurity Defenders

The project’s initial partners, who build and maintain software essential to the functioning of the internet and other vital infrastructure, report that these discoveries significantly reduce risk for billions of end users. This demonstrates how AI can be a powerful ally in the race for security.

What Lies Ahead

Project Glasswing is just beginning. As patches for the vulnerabilities discovered by Mythos Preview are being deployed, more comprehensive details about the findings will be shared. The team is also exploring how Mythos-class models might be released in the future.

Conclusion: A New Era of Security

The early successes of Project Glasswing demonstrate the potential of AI to transform cybersecurity. As threats continue to evolve, initiatives like this are crucial to safeguarding our digital infrastructure.

Let's discuss your project in 15 minutes.

cybersécurité vulnérabilités critiques intelligence artificielle Mythos Preview Project Glasswing
Deepthix newsletter · 100% AI · every Monday 8am

An AI agent reads tech for you.

Our AI agent scans ~200 sources per week and ships the best articles to your inbox Monday 8am. Free. One click to unsubscribe.

Visit the newsletter page →

Want to automate your operations?

Let's talk about your project in 15 minutes.

Book a call