Anthropic has significantly broadened the scope of Project Glasswing, its AI-powered cybersecurity initiative, by extending access to Claude Mythos Preview to approximately 150 new organizations spanning more than fifteen countries.
The expansion follows several weeks of close collaboration with existing partners, the security industry, open-source software maintainers, and the US government.
Launched in early April, Project Glasswing was introduced as a collaborative effort to secure the world’s most critical software infrastructure.
The initial cohort of roughly 50 partner organizations was granted access to Claude Mythos Preview, Anthropic’s frontier AI model with advanced cybersecurity capabilities to scan their codebases for vulnerabilities at scale.
Anthropic Expands Project Glasswing
In just the first weeks of deployment, these initial partners collectively discovered more than 10,000 high- or critical-severity security flaws across their codebases, demonstrating the model’s capacity to surface vulnerabilities at a speed and scale previously unattainable through traditional security tooling.
The 150 newly added organizations represent a deliberate expansion into industries that were underrepresented in the first cohort, including power, water, healthcare, communications, and hardware.
Many of these partners are vendors and nonprofits maintaining codebases relied upon by governments and organizations worldwide.
A key criterion for inclusion is impact: a successful cyberattack on any partner’s codebase could affect more than 100 million people, with significant implications for national and global security.
Each new partner must meet Anthropic’s security requirements before gaining access to Mythos Preview.
Anthropic has issued a pointed warning alongside the expansion. Within 6 to 12 months, the company expects multiple AI providers to release Mythos-class models, potentially without the safeguards needed to prevent misuse.
This scenario could dramatically increase the frequency and unpredictability of cyberattacks, putting critical infrastructure at unprecedented risk.
“Cheap, fast AI models with powerful cyber capabilities are around the corner,” Anthropic stated, framing Project Glasswing as a proactive measure to ensure defenders can keep pace with rapidly evolving offensive capabilities.
The company sees its role as twofold: helping the software industry adapt by safely providing wide access to better models and tools, and steadily shifting support from finding vulnerabilities to disclosing, fixing, and deploying patched software.
To support the broader security community, Anthropic is releasing, on request, to trusted security teams the tools developed to help Project Glasswing partners find vulnerabilities faster.
Additionally, the company has launched Claude Security, a product built on frontier public models like Claude Opus 4.8 that scans codebases and suggests patches for identified weaknesses. Partners are already using Mythos Preview well beyond initial vulnerability discovery.
They are writing and validating security patches, running pre-release code checks to prevent new vulnerabilities, conducting penetration testing using simulated cyberattack scenarios, automating threat detection and response workflows, and even rebuilding legacy codebases in memory-safe languages to eliminate entire classes of vulnerabilities at the source.
Anthropic acknowledges that the current bottleneck in cybersecurity is no longer finding vulnerabilities, but verifying, disclosing, and patching them.
The company is in active discussions with third parties about scaling up vulnerability remediation in open-source software and developing standardized practices for disclosing findings to open-source maintainers, making reports easier to triage and act upon.
Anthropic plans to continue expanding Project Glasswing, prioritizing essential infrastructure providers, critical open-source software maintainers, and safety testers across both US and international organizations.
The company also intends to scale its Cyber Verification Program, which would grant Mythos-class capabilities to additional organizations for specific cyberdefense tasks.
Longer-term, Anthropic aims to make Mythos-level capabilities available for general access contingent on developing sufficiently robust safeguards that prevent misuse while preserving defensive utility.
The company views Project Glasswing as a foundational model for how the AI industry should respond when models cross critical capability thresholds, with the ultimate goal of establishing a permanent structural advantage for defenders over attackers.
Follow us on Google News , LinkedIn and X to Get More Instant Updates. Set Cyberpress as a Preferred Source in Google.
