Anthropic Says Mythos Has Uncovered Over 10K Vulnerabilities
Anthropic Says Mythos Has Uncovered Over 10K Vulnerabilities
Publish Date: 2026-05-25 20:04:00
Source Domain: www.pymnts.com
- Anthropic’s Mythos Preview model identified over 10,000 cybersecurity vulnerabilities as part of Project Glasswing.
- Project Glasswing is a collaboration between Anthropic and 50 partner organizations aimed at using AI to counter AI-powered cyberattacks.
- Participants in Project Glasswing reported a tenfold increase in their bug-finding rates, with Cloudflare specifically finding 2,000 bugs including 400 high- or critical-severity ones.
- Anthropic warns that without adequate safeguards, similarly capable AI models could easily be used to exploit software vulnerabilities.
- Research indicates that large enterprises face heightened risks of AI-powered cyberattacks, including deepfakes and automated data scraping, with 58% of firms over $1 billion in revenue reporting related attacks in the past year.
