AI-found bugs aren't proving any easier to exploit despite the hype

VulnCheck analyzed 1,061 publicly attributed AI-assisted vulnerability discoveries from Anthropic's Project Glasswing and the Berkeley Vulnerability Research Initiative, cross-referencing them against its Known Exploited Vulnerability database, and found only 14 (1.3 percent) confirmed exploited in the wild — nearly identical to the baseline rate for all vulnerabilities. The research concludes AI mainly increases the volume of flaws researchers uncover rather than the share attackers weaponize, suggesting claims that frontier AI dramatically favors attackers are overhyped.