AI detected critical flaws in the foundations of the Web: how an invisible error threatened much of the digital world

For over a decade, two lines of code remained hidden in a programming language used by 70% of websites, posing a significant security risk. Advanced AI models from OpenAI and Google independently discovered vulnerabilities, leading to urgent global responses from cybersecurity agencies. The flaws, classified as CVE-2026-109618 and CVE-2026-109619, allow remote attackers to execute malicious code without authentication. This incident highlights the effectiveness of AI in identifying security issues but also raises concerns about the speed of human response to such threats. The need for improved organizational capacity to address vulnerabilities is critical as many servers still run outdated software.