Anthropic brings Mythos 5 to its Claude Security vulnerability scanner

Sicherheitsmeldung mit Schweregrad noch nicht bewertet. Technische Details im Tab „Originaltext“; empfohlene Schritte in der Checkliste.

Originaltext der Quelle (nicht übersetzt) – die deutsche Zusammenfassung findest du im Tab „Kurzfassung".

Earlier this year, Anthropic launched Claude Security, an enterprise tool that helps development teams scan their codebase for security vulnerabilities and patch them. On Friday, the company gave Claude Security a major upgrade by bringing its Claude Mythos 5 model to the service.

In addition, Anthropic is also working with other cybersecurity companies to help them integrate Mythos 5 into their products, and it is launching a new Defender Advantage Fund (0xDAF) that will provide $35 million in credits to find and patch vulnerabilities in open source software.

Anthropic previously also launched its Cyber Verification Program, which lets vetted defenders run dual-use cybersecurity work on Opus and Sonnet with fewer blocks. Those organizations will also get safeguarded access to Claude Mythos soon, Anthropic says.

The Mythos 5 story: too good to launch

Mythos 5, of course, is the model that was so good in performing tasks in high-risk domains that Anthropic didn’t give it a public release. Instead, in June, it launched Fable 5, which is essentially Mythos 5 but with very strict guardrails.

Mythos 5, at the time, was only available to about 150 partners in Anthropic’s Project Glasswing program (and there’s the whole sidestory of Fable 5 getting banned and unbanned by the U.S. government).

Mythos 5 in Claude Security

So why does the company feel comfortable adding Mythos 5 to Claude Security now?

“The riskiest behavior occurs when a user has direct access to a model, where a malicious actor can try to steer it toward harmful uses,” Anthropic writes in the …

Vollständiger Originalartikel ↗

• Betroffenheit im eigenen Stack prüfen: Versionen/Komponenten abgleichen.
• Originalquelle / Hersteller-Advisory lesen: https://thenewstack.io/anthropic-mythos-claude-security/
• Verfügbaren Patch oder Workaround einspielen und dokumentieren.