Anthropic is widening access to the cybersecurity capabilities of Claude Mythos 5, its most capable model, while still keeping the model itself out of direct reach. In a post on its site, the company said Mythos 5 now powers Claude Security scans, is coming to partners’ cyber defense tools, and will anchor a new $35 million fund for open-source security.
Mythos has been tightly held since April, when Anthropic launched Project Glasswing to give a small group of organizations securing critical software early access to Claude Mythos Preview and then Mythos 5. The idea was to give defenders time to find and fix vulnerabilities before similar capabilities became widely available or reached attackers. Claude Fable 5 later made a Mythos-class model broadly available, with dual-use cyber work blocked.
Outputs, Not the Model
The new approach rests on a distinction Anthropic draws between direct model access, where a malicious user can try to steer the model toward harm, and products that return only a specific artifact such as a patch or a security alert. “The riskiest behavior occurs when a user has direct access to a model,” the company wrote.
Anthropic is working with its cybersecurity technology and services partners to build Mythos 5 into the tools security teams already use for security operations, incident response, threat intelligence and detection engineering. End users of those products will not interact with Mythos directly. A remediation tool, for example, might return a list of suggested patches generated by Mythos, but the user could not prompt the model to write an exploit.
Claude Security Scans Run on Mythos 5
Claude Security, in public beta for Claude Enterprise customers, now runs its codebase scans on Mythos 5. Each finding comes back with a CWE category, confidence and severity ratings, and a suggested fix, according to SecurityWeek. Fixes are implemented through Claude Code on the web, and every patch must be reviewed and approved by a human. Scans are billed as standard token usage under existing plans.
$35 Million for Open Source
The new Defender Advantage Fund (0xDAF) will provide $35 million in Claude credits to organizations helping open-source maintainers. Grants will target patching live vulnerabilities in widely used projects, automating scanning and patching in ways other projects can reuse, and security approaches that resist whole classes of attack. Anthropic is starting with a small number of larger pilot grants. Through Project Glasswing it had already given $4 million in direct donations to open-source security organizations.
Anthropic also plans to expand its Cyber Verification Program, which gives vetted defenders reduced safeguards on Opus and Sonnet models, to cover broader dual-use capabilities such as vulnerability triage and validation, with Mythos-class access to follow in the coming weeks.
For security teams, the practical change is that Mythos-grade vulnerability scanning is now available inside an existing Enterprise plan and, soon, inside familiar vendor products, without the risk that comes with handing out an unrestricted offensive-capable model.


WebProNews is an iEntry Publication