Anthropic is expanding defender access to its most capable cybersecurity-oriented AI models, pairing new partner integrations with an updated Claude Security beta now running on Mythos 5 and launching a $35 million open source funding program. The announcement builds on Project Glasswing, which began in April by giving a small group of organizations early access to Claude Mythos Preview and its successor, Mythos 5.
How Anthropic is broadening Mythos 5 access
The latest changes are built around a core idea: give defenders more useful output from Mythos-class models while keeping direct, unrestricted interaction with the models restricted. Anthropic has described direct, unrestricted access as the riskiest scenario, noting that the risk drops sharply when users instead receive specific defensive outputs such as a patch or a security alert.
Partner integrations for security operations
Anthropic is working with cybersecurity partners to embed Mythos 5 inside the security operations, incident response, and detection tools that teams already use to protect hospitals, utilities, financial systems, and the software supply chain. End users will not interact with Mythos directly. Instead, they will work through purpose-built interfaces that run the model in the background and return only a defined output, such as a list of suggested patches, with abuse-prevention checks meant to keep the model within that scope.
Claude Security now scans on Mythos 5
Claude Security, currently in public beta for Claude Enterprise customers, now runs its codebase scans on Mythos 5. Each surfaced finding includes a CWE category, confidence and severity ratings, and a suggested fix. Any fix still has to be implemented through Claude Code and approved by a human before deployment.
What is the $35M open source cybersecurity fund?
Anthropic is launching a $35 million open source funding program aimed at strengthening the open source software ecosystem against cyber threats. The fund is part of the company’s wider effort to extend the reach of its AI-powered defensive capabilities beyond its own products, backing projects that help maintainers and defenders find and remediate vulnerabilities in widely used open source components.
Why Project Glasswing matters for defenders
Project Glasswing, launched in April, gave a small group of organizations early access to Claude Mythos Preview and its successor, Mythos 5. Anthropic said the goal was to give defenders time to find and fix vulnerabilities before comparable capabilities became widely available or fell into the hands of malicious actors. Claude Fable 5 followed as a broadly available model that keeps dual-use cyber work blocked.
The expanded Cyber Verification Program
Alongside the integrations and the open source fund, Anthropic plans to expand its Cyber Verification Program. The program is designed to validate that defensive uses of Mythos-class models stay within safe scopes and that partners embedding the models into their tools do so responsibly, with abuse-prevention controls in place.
What this means in practice
For security teams, the practical effect is a wider set of pathways to put Mythos-class capabilities to work without gaining direct, unrestricted access to the underlying models. Defenders using Claude Enterprise will see richer codebase scans on Mythos 5, with CWE-tagged findings and severity ratings. Teams relying on third-party security operations tools will see Mythos 5 reach them through vendor integrations, with output constrained to specific defensive tasks. Open source maintainers and projects that secure critical infrastructure will have a new funding channel for defensive work.
FAQ
What is Mythos 5?
Mythos 5 is the successor to Claude Mythos Preview, an advanced AI model that Anthropic has been developing with cybersecurity in mind. It is the model that powers the new partner integrations and the updated Claude Security codebase scans.
How is Anthropic keeping Mythos 5 restricted?
End users do not interact with Mythos 5 directly. They access its capabilities through purpose-built interfaces that run the model in the background and return only a defined output, such as a list of suggested patches, with abuse-prevention checks meant to keep the model within that scope. Claude Fable 5, the broadly available successor model, keeps dual-use cyber work blocked.
What does the $35M open source fund cover?
Anthropic is launching a $35 million open source funding program to strengthen the open source software ecosystem against cyber threats, backing projects that help maintainers and defenders find and remediate vulnerabilities in widely used open source components.
This article summarizes reporting from securityweek.com. See our editorial disclaimer for how our articles are produced.
Run a free scan to see your AI Visibility Score, SEO rating, and local citation accuracy.