Hasty Briefsbeta

Bilingual

Six curl CVEs after OpenAI and Anthropic came back with zero

2 months ago
  • AISLE discovered six curl CVEs within days after OpenAI and Anthropic's frontier AI systems reported zero findings.
  • The six vulnerabilities were rated Low severity and fixed in curl 8.22.0, with AISLE credited as the reporter.
  • The comparison was unusually clean because Daniel Stenberg published the zero results before AISLE ran its analysis.
  • Curl maintainers independently validated and assigned CVEs, providing strong external validation for zero-day discovery.
  • Linux stable maintainer Greg Kroah-Hartman noted a similar pattern with AISLE outperforming other AI systems on Linux.
  • This result supports AISLE's 'System over Model' thesis, demonstrating specialized AI systems can outperform frontier AI labs at real-world zero-day discovery.