What's Happening?
Wiz Research, in collaboration with Ikotas Labs, Metnir, Rapid7, and Xint, has independently discovered two critical unauthenticated remote code execution vulnerabilities in NGINX, tracked as CVE-2026-123633 and CVE-2026-123007. NGINX is a widely used
open-source web server that powers over 225 million websites, often acting as a reverse proxy for web applications and APIs. These flaws, rated with a CVSS 4.0 score of 9.2, are present in NGINX itself, meaning that even applications with secure code can be exposed. Both vulnerabilities require HTTP/2 to be enabled on the affected NGINX server. CVE-2026-123007, an integer underflow vulnerability, has existed since NGINX version 1.9.5 in 2015, while CVE-2026-123633, a use-after-free vulnerability, has been present for nearly six years. Wiz Research utilized AI code scanning solutions, including Wiz Atlas, Gemini 4 Carbon, and ChatGPT 5.6 Sol, to discover and validate the impact of these vulnerabilities. The findings were reported to F5, the company behind NGINX, leading to a coordinated disclosure and the release of security updates.
Why It's Important?
The discovery of these critical vulnerabilities in NGINX carries significant implications for cybersecurity across the U.S. and globally. Given NGINX's pervasive use as a front-door for modern web architectures, successful exploitation could allow attackers to gain control of affected servers, steal sensitive data, tamper with websites, or disrupt critical services. This poses a substantial risk to businesses, government entities, and any organization relying on NGINX for their web infrastructure. The fact that these flaws have remained undetected for years, some for over a decade, highlights the evolving landscape of vulnerability research, where AI is now playing a crucial role in uncovering complex memory-safety issues. This shift means that long-standing software, previously considered robust, may now be susceptible to new forms of analysis, increasing the urgency for organizations to regularly update and patch their systems. The potential for widespread impact across multiple applications served by a single NGINX deployment amplifies the severity of these findings, making prompt remediation essential to prevent large-scale data breaches and service interruptions.
What's Next?
Organizations utilizing NGINX are strongly advised to identify affected deployments with HTTP/2 enabled and apply the security updates immediately. F5 has released patched versions, including NGINX Open Source 1.30.7 stable, 1.31.9 mainline, and corresponding NGINX Plus versions. If immediate upgrading is not feasible, a temporary mitigation involves disabling HTTP/2 on affected NGINX instances by setting `http2 off;` in server blocks and removing the `http2` parameter from listen directives. Security teams should also monitor for unexpected NGINX worker crashes or unusual commands launched by NGINX, as these could be general warning signs of compromise. The broader implication is that the cybersecurity community will likely see an increased focus on AI-driven vulnerability research, leading to the discovery of more long-standing flaws in critical infrastructure. This will necessitate a more proactive and rapid response from software vendors and security teams to patch and secure their systems against newly identified threats.
Beyond the Headlines
The uncovering of these NGINX vulnerabilities by AI-powered tools signals a profound shift in the economics and capabilities of vulnerability research. Historically, memory-safety flaws like these were exceptionally difficult to discover and exploit due to their complexity, requiring intricate tracing of memory operations and code paths. AI is fundamentally dissolving this barrier, enabling both defenders and attackers to uncover deeply embedded flaws in software that underpins the internet. This development raises ethical considerations regarding the responsible use of AI in cybersecurity, as the same tools that help researchers can also be leveraged by malicious actors. The fact that multiple independent research teams, some using AI, discovered these vulnerabilities after they had persisted for years underscores that complexity can no longer be relied upon as a security measure. This necessitates a re-evaluation of software development and auditing practices, emphasizing continuous security assessments and the adoption of advanced tools to keep pace with evolving threat landscapes. The long-term implication is a continuous arms race between AI-powered security and AI-powered attacks, demanding constant innovation in defensive strategies.













