AI-Assisted Vulnerability Discovery Outpaces Remediation Capacity
3 sources · Updated 3d ago · First seen Apr 15, 2026

AI-assisted security research is increasing the volume of actionable vulnerability reports faster than organizations can resolve them. HackerOne reported a record 46,947 submissions in March 2026, up 76% year over year, while the share validated as exploitable remained roughly 25%. Critical and high-severity issues represented 32% of validated findings, above the historical 26–28% baseline. Remediation throughput rose only 19%, pushing backlogs to a record high even as median remediation time for critical findings fell below 15 days. HackerOne attributed the growing volume and severity to human researchers using AI to broaden testing and uncover more complex exploit chains.
The pressure is also reaching open-source maintainers. Curl founder Daniel Stenberg said low-quality AI-generated security reports had disappeared from the project in recent months, replaced by an unprecedented flow of strong reports, almost all produced with AI assistance. He said the volume placed maintainers under serious load, although many submissions were ultimately classified as non-security bugs. Together, these accounts highlight a capacity problem extending beyond report quality: validation and engineering resources must keep pace with discovery. HackerOne recommends continuous vulnerability intake, scalable validation, remediation integrated into engineering workflows, and systemic fixes for recurring vulnerability classes rather than relying solely on faster handling of individual critical issues.


