tech
Anthropic's Claude Mythos found 10,000 critical vulnerabilities in one month. The patches can't keep up.
Anthropic’s Glasswing project found 10,000+ critical flaws across 1,000 open-source projects in a month. Only 97 have been patched.
TL;DR
- Anthropic's Glasswing project discovered over 10,000 critical vulnerability candidates in 1,000+ open-source projects in one month.
- Only 97 of the confirmed critical flaws have been patched, highlighting a significant gap between discovery and remediation.
- Claude Mythos Preview, an AI model used by Glasswing, can identify vulnerabilities at a pace the open-source ecosystem cannot absorb.
- A critical flaw in WolfSSL, a widely used embedded TLS library, was a notable finding, with potential for attackers to forge certificates.
- Access to Claude Mythos Preview is restricted to approximately 50 organizations described as "systemically important cyber defenders."
- The AI model has also been used for defensive purposes, such as detecting and preventing a $1.5 million fraudulent wire transfer.
- Anthropic and OpenAI are developing specialized AI models for cybersecurity, restricting public access due to concerns about misuse.
- The rapid AI advancements necessitate a call for developers to shorten patch cycles and for organizations to harden security measures.