AI accelerates vulnerability discovery, ending the myth of security through obscurity

Artificial-intelligence models are being used by software vendors and independent researchers to hunt for vulnerabilities, revealing obscure and legacy flaws in everything from old Unix protocols to modern Microsoft components. This surge has led to record-breaking counts of disclosed CVEs and a mounting backlog for maintainers. Security officials note that attackers are equally leveraging AI to reverse-engineer patches and launch exploits within hours, including against critical infrastructure like Siemens PLCs.

Interviews with former US cyber officials underscore the risk to operational technology, where AI can compensate for a lack of specialized expertise. Despite the flood of findings, AI-generated patches succeed in only about a quarter of cases and often alter application behavior, according to 1Password and Veracode research. Experts argue that organizations must shift focus from sheer bug counts to process improvements and proactive security design.