Anthropic on June 2 announced that it is expanding Project Glasswing, its cybersecurity initiative to secure critical software, to approximately 150 additional organisations across more than 15 countries.
The expansion follows the company’s initial rollout in April, when about 50 partners were granted access to Claude Mythos Preview, an AI model trained to identify software vulnerabilities. Anthropic recently said those partners had discovered more than 10,000 high- and critical-severity security flaws using the system.
The new group includes organisations from sectors such as power, water, healthcare, communications and hardware, areas that were less represented in the initial cohort. Many of the new participants are software vendors and nonprofit maintainers whose codebases are used by governments and organisations worldwide.
“What each partner has in common is that a successful attack on their codebase could be catastrophic,” Anthropic said. The company added that for most partners, “a major attack could affect more than 100 million people, with important ramifications for both global and national security.”
Anthropic said the expansion is part of its broader effort to improve software security through AI while helping the cybersecurity industry adapt to increasingly capable AI systems.
The company warned that “cheap, fast AI models with powerful cyber capabilities are around the corner” and reiterated its expectation that many AI developers could have models with capabilities comparable to Claude Mythos Preview within six to 12 months.
“Mythos Preview continues a long-term trend that we’ve been warning about for some time,” Anthropic said, adding that some future models could be released without safeguards that prevent misuse, potentially increasing the frequency and unpredictability of cyberattacks.
As part of its efforts to support defenders, Anthropic said it recently launched Claude Security, a product that uses its public frontier models, including Claude Opus 4.8, to scan codebases and suggest patches. The company is also making available, upon request, to trusted security teams, tools developed for Project Glasswing participants to help identify vulnerabilities more quickly.
Anthropic said the cybersecurity challenge is shifting from vulnerability discovery to the processes of verification, disclosure and patching. Many Project Glasswing partners are already using Mythos Preview to generate software patches and conduct pre-release security checks, according to the company.
The model can also be applied to penetration testing, threat detection and response, and modernising legacy software by rebuilding codebases in memory-safe programming languages, Anthropic said.
The company said it is discussing ways to scale vulnerability review and patching efforts in open-source software and is working on improving vulnerability disclosure practices for open-source maintainers.
Looking ahead, Anthropic said it plans to further expand Project Glasswing, focusing on critical infrastructure providers, maintainers of essential open-source software and safety testers. It also intends to broaden its Cyber Verification Program, which would provide Mythos-class capabilities to more organisations for specific cyber defence tasks.
“We’re working as quickly as we can to safely release Mythos-level capabilities in general access,” Anthropic said, while acknowledging that safeguards capable of preventing misuse of advanced cyber models remain an unresolved challenge across the AI industry.
ALSO READ: Alteryx Inspire 2026: Three Questions Every Data Leader Should Take to Orlando
