Anthropic Expands Mythos Access to 150 More Organizations

https://www.infosecurity-magazine.com/news/anthropic-glasswing-expansion/

Publish Date: 2026-06-12 04:37:16

Source Domain: www.infosecurity-magazine.com

Summary:

Anthropic has significantly expanded its Project Glasswing, which utilizes their leading AI model to identify security vulnerabilities in critical software systems. Since its launch, an initial group of 50 partners has unearthed over 10,000 high- or critical-severity flaws, indicating the model’s effectiveness. To further bolster security, Anthropic is extending access to 150 more organizations across more than 15 countries, representing sectors such as power, water, healthcare, communications, and hardware. This is crucial as any breach in these sectors could catastrophically impact over 100 million people. However, the issue remains that discovering vulnerabilities is outpacing their remediation. Anthropic warns that this gap could be exacerbated as competitors develop similar AI models without adequate safeguards. This disparity highlights the urgent need for industries to improve their software security infrastructure to validate, prioritize, and fix these flaws rapidly, suggesting a shift towards integrating AI in threat modeling and secure design over traditional methods.

Key Points:

Anthropic has extended access to its AI model, Claude Mythos Preview, to 150 additional organizations to find critical software vulnerabilities.
These new partners cover sectors previously under-represented, such as healthcare and communications, and are located in more than 15 countries.
Anthropic emphasizes that vulnerability discovery with AI is outpacing the ability to fix these vulnerabilities, creating a potential security bottleneck.
The firm anticipates that rival developers could soon deploy powerful AI models without mitigation strategies, exacerbating the issue.
Experts argue for leveraging AI toward threat modeling and secure design rather than relying on the obsolete scan-and-patch approach to improve software security.