Two-Year PHP Bug Turned Critical as AI Researchers Exploit It in Days

The central tension in this latest cybersecurity development lies in the shifting balance of power between human developers and artificial intelligence. For two years, a critical flaw in PHP, the open-source language powering an estimated 70% of the web, remained a puzzle that stumped human experts. However, recent events demonstrate that AI-assisted researchers can turn such dormant vulnerabilities into working exploits in mere weeks, fundamentally altering the timeline for digital defense.

The scale of the issue is significant. The PHP Foundation released patched versions 8.2.35, 8.3.36, 8.4.27, and 8.5.12 on October 21, 2026, moving up the original October 22 schedule after premature disclosures sparked public concern. Both vulnerabilities, identified as CVE-2026-109618 and CVE-2026-109619, carry a critical severity score of 9.2 out of 10. These flaws reside in the core engine, meaning websites are vulnerable even if their specific application code is secure, and they can be exploited remotely without login credentials.

The discovery process highlights the role of specific organizations and tools. Researchers from Calif, a California-based firm and founding partner of OpenAI's 'Patch the Planet' initiative, began auditing PHP in late July 2026. On August 11, Calif researcher Ryan A. and OpenAI researchers using the GPT-5.6 Sol model independently identified the exploitable nature of the flaw. Another firm, Wiz, also identified an additional vulnerability. OpenAI coordinated a coalition of maintainers and government agencies to fix the issues over a two-month period, a timeline attributed to the complexity of the required engine changes.

The technical nature of the bugs involves errors in memory management and error handling. One flaw occurs when PHP runs out of allocated memory, potentially leaving incomplete data that the engine later treats as valid. The other involves a fatal error that leaves PHP pointing to freed memory, allowing subsequent requests to corrupt it. Researchers demonstrated that these weaknesses could expose sensitive data, including passwords and API keys, and allow attackers to execute malicious code on affected servers.

The stakes extend beyond a single patch, touching on the broader concept of the 'Defender's Window.' OpenAI argues that there is a narrow timeframe where organizations can use advanced AI to patch defenses before malicious actors gain similar capabilities. This perspective challenges the view that AI-only vulnerabilities are harmless if humans cannot exploit them. The company points to recent findings by Anthropic’s Claude Mythos, which identified thousands of zero-day vulnerabilities across major operating systems and browsers, suggesting that proactive AI-driven defense is now a collective responsibility.

For the national audience, this incident underscores the ubiquity of the technology underlying the modern web. While PHP 8.1 and older versions will not receive official fixes, the coordinated response aims to ensure that the majority of sites can upgrade safely. The event serves as a case study in how AI is accelerating both the discovery and exploitation of security flaws, forcing the industry to adapt its defensive strategies to match the new speed of attack.