AI accelerates discovery rate, especially for cyber vulnerabilities · METR analysis

A new analysis from METR suggests that the rate of publicly disclosed discoveries has accelerated, particularly in the realm of cybersecurity vulnerabilities. While AI's contribution is evident in some areas, such as a sharp increase in AI-marked disclosures for projects like cURL and OpenSSL, its impact on the overall discovery rate is harder to definitively quantify. The study notes that while AI has doubled arXiv submissions in some fields, its role in solving complex mathematical problems or improving optimization techniques shows less dramatic acceleration, with higher-severity vulnerabilities exhibiting lower growth rates. AI

IMPACT Suggests AI is accelerating discovery, particularly in cybersecurity, though its broader impact is still being quantified.

RANK_REASON Analysis of existing data on discovery rates, not a primary research release or product launch.

Read on METR (Model Evaluation & Threat Research) →

• arXiv
• Cisa
• curl
• Firefox
• Greg Burnham
• Microsoft
• National Vulnerability Database
• OpenSSL
• Vulncheck KEVs

• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →