WASHINGTON, D.C. — Anthropic announced on Tuesday that 150 new partners across more than 15 countries will gain access to its Mythos artificial intelligence model as part of an expansion of Project Glasswing. The move broadens the initiative to include industries underrepresented in the initial launch, such as power, water, healthcare, communications, and hardware.
The Mythos model, first introduced in April as a restricted system available only to select companies and security researchers, has demonstrated capabilities in identifying software vulnerabilities. Since the project’s initial testing phase with 50 partners began in April, participants have revealed more than 10,000 high or critical security flaws. Mythos identified more than 23,000 potential vulnerabilities, with over 6,000 estimated to be severe.
New partners must meet security requirements before receiving access to the model. Many of the newly added organizations are vendors and maintainers of widely used codebases relied upon by governments and institutions globally. According to the Financial Times, these include Okta, Samsung, the EU cybersecurity agency ENISA, and NATO. A successful cyberattack targeting the products of most new participants could affect more than 100 million people and carry national and global security implications.
Anthropic did not disclose the full list of new companies joining the coalition. Organizations such as Mozilla, Palo Alto Networks, and Cloudflare reported positive results when using Mythos against their own products. Of the thousands of security bugs found, only 75 critical and high-severity issues have been patched so far.
Anthropic said Mythos can assist not only in finding vulnerabilities but also in verification and patching. The company is working with others to “substantially scale up the reviewing and patching of vulnerabilities in open-source software.” It added that it is developing best practices for disclosing vulnerabilities to open-source maintainers to make reports easier to triage and act upon.
“This expansion is the next step toward our long-term goals: for AI to make all software more secure, and for us to help the industry adjust to how AI could change many of the core assumptions of cybersecurity,” Anthropic said in a blog post. The company confirmed plans to release Mythos-class models to the general public in the coming weeks but has not set a specific date.
No independent assessment was available for this report.