AI-Discovered Security Flaws Are Not Yet Translating Into More Exploitation
Security Boulevard, Tuesday, July 28th, 2026
Only 1.3% of AI-discovered vulnerabilities have been exploited, suggesting defenders currently benefit more than attackers.
Despite fears that frontier AI models would enable widespread exploitation, data shows AI-discovered vulnerabilities are not fueling cyberattacks at higher rates than traditionally discovered flaws.
VulnCheck researcher Patrick Garrity analyzed over 1,000 AI-discovered vulnerabilities and found that just 14 (1.3%) were exploited in the wild, matching historical exploitation rates.
The data does not suggest that AI-discovered vulnerabilities are inherently more likely to be exploited than those found through traditional methods. Currently, defenders appear to have the advantage in leveraging frontier AI models to identify and patch flaws before attackers can exploit them, though researchers caution this advantage may shift as more powerful models become widely available.