Anthropic’s Claude Mythos AI Uncovers Decades-Old Software Vulnerabilities
- Claude Mythos scored an impressive 83.1% on Cybergym, significantly outperforming its predecessor Claude Opus.
- The AI model identified thousands of zero-day vulnerabilities across major operating systems and browsers.
- A notable discovery was a 27-year-old OpenBSD flaw, found autonomously after approximately $20,000 worth of testing.
- Project Glasswing launched with $100 million in Mythos usage credits for defenders and includes partners like Amazon Web Services and Google.
- Less than 1% of the vulnerabilities identified have been fully patched to date.
Claude Mythos AI has demonstrated unprecedented capability in identifying critical software vulnerabilities that have eluded human experts for decades, highlighting the potential for AI to enhance cybersecurity defenses significantly.
Despite its achievements, only a small fraction of discovered flaws have been patched, underscoring the need for rapid response mechanisms in cybersecurity frameworks.( Source)
