Anthropic is expanding how security teams can tap into the cyber capabilities of its advanced AI models, rolling out partner integrations, an updated Claude Security offering, a new open source funding program, and plans to grow its Cyber Verification Program.
The effort builds on Project Glasswing, launched in April, which gave a small set of organizations early access to Claude Mythos Preview and its successor, Mythos 5. Anthropic says the intent is to let defenders find and fix vulnerabilities before similar capabilities become widely available or end up in the hands of attackers. Claude Fable 5, a broadly available model, keeps dual-use cyber functionality blocked entirely.
Output-only access, not direct interaction
Anthropic’s approach centers on limiting direct, unrestricted access to Mythos-class models, which it considers the riskiest scenario. Instead, users receive specific defensive outputs, such as a patch recommendation or a security alert, generated by the model running in the background.
Under the new integrations, cybersecurity partners will embed Mythos 5 into security operations, incident response, and detection tools already deployed across hospitals, utilities, financial systems, and software supply chains. End users interact only with purpose-built interfaces that return defined outputs, such as suggested patches, with abuse-prevention checks meant to keep the model within scope.
Claude Security now runs on Mythos 5
Claude Security, in public beta for Claude Enterprise customers, now performs its codebase scans using Mythos 5. Findings come tagged with a CWE category, confidence and severity ratings, and a suggested fix. Any proposed fix still must be implemented through Claude Code and approved by a human before deployment, keeping the model itself inaccessible even as its analysis reaches defenders.
$35 million for open source security
Anthropic also launched the Defender Advantage Fund (0xDAF), committing $35 million in Claude credits to organizations helping open source maintainers secure their projects. This follows roughly $4 million in direct donations and support already delivered through Project Glasswing, including coordinated efforts known as Akrites and Gold Eagle. Grants will target patching live vulnerabilities, building reusable scanning and patching workflows, and developing approaches meant to resist entire classes of attack, starting with a small number of larger pilot grants.
Cyber Verification Program expansion
Anthropic’s Cyber Verification Program, which already grants vetted organizations reduced safeguards on Claude Opus and Sonnet for authorized security work, will expand in the coming weeks to cover broader dual-use capabilities including vulnerability triaging and validation, with Mythos-class access to follow. Anthropic is also continuing to expand Mythos access through Project Glasswing alongside US government partners, focused on critical infrastructure organizations meeting strict security control requirements. Security teams can apply to the Cyber Verification Program now, with further rollout details expected soon.