PHP Security: AI Models Expose Two-Year-Old Critical Flaws

OpenAI announced on October 21 that its artificial intelligence models played a pivotal role in identifying two critical security weaknesses in PHP, the open-source language that powers an estimated 70% of the internet. This discovery was made by researchers participating in the company's Patch the Planet cybersecurity initiative, who utilized the GPT-5.6 Sol model to analyze the codebase. The findings prompted a rapid coordination effort involving PHP maintainers, security providers, and government agencies to develop and distribute fixes. OpenAI described the event as one of the most significant cybersecurity discoveries of 2026, largely due to the widespread reliance on PHP across global web infrastructure.

The urgency of the situation was driven by the severity of the bugs and the speed at which they could be weaponized. Both vulnerabilities were assigned a critical severity score of 9.2 out of 10. The flaws allow remote attackers to execute malicious code on servers without needing login credentials, potentially giving them control over websites. Furthermore, the vulnerabilities can expose sensitive data stored in server memory, including passwords, authentication tokens, and API keys. Because the issues reside in the core PHP engine, any site using the language is at risk, even if the specific application code is otherwise secure.

The technical details of the two flaws involve errors in how PHP manages computer memory and processes errors. The first vulnerability, tracked as CVE-2026-109618, occurs when the system runs out of allocated memory during a request. This can leave behind incomplete internal data that the engine mistakenly treats as valid in later operations. The second flaw, CVE-2026-109619, involves a fatal error that leaves the system pointing to memory that has already been freed. This allows subsequent requests to access or corrupt that memory, creating a pathway for exploitation. These core engine issues mean that the vulnerability is inherent to the language's infrastructure rather than specific to individual applications.

The timeline of the discovery highlights the changing dynamics of software security. The PHP Foundation had been aware of the underlying flaw since 2024 but struggled to find a method to exploit it or a way to fix it. Human developers and security teams examined the code for two years without a full resolution. In contrast, AI-assisted researchers demonstrated practical exploitation methods within just two weeks. This stark difference in speed underscores the shift in how vulnerabilities are found and addressed, moving from a multi-year human audit process to a rapid AI-driven analysis.

The audit process began on July 30, 2026, at the request of the PHP Foundation. Researchers from Calif, a California-based cybersecurity firm and the founding partner of the Patch the Planet program, started using OpenAI tools to examine the code. On August 11, a Calif researcher named Ryan A. identified the first exploitable vulnerability. OpenAI researchers confirmed this finding on the same day. Shortly after, another cybersecurity firm, Wiz, independently identified the second vulnerability. This collaborative effort brought together multiple entities to ensure the issues were thoroughly understood before a fix was attempted.

Following the identification of the flaws, a coalition worked for two months to implement the necessary patches. OpenAI attributed the lengthy timeline to the complexity of the task, which required major changes to the PHP engine itself. The PHP Foundation eventually released patched versions 8.2.35, 8.3.36, 8.4.27, and 8.5.12. It is important to note that PHP 8.1 and older versions will not receive official fixes. The coordinated release was originally scheduled for October 22, but it was moved forward to October 21 after premature disclosures on October 20 led to public concern and outcry.

OpenAI is using this incident to illustrate a concept it calls the Defender's Window. Introduced after its models hacked Hugging Face, this concept describes a narrow timeframe where organizations can use advanced AI models to find and patch defenses before those same capabilities become widely available to malicious actors. OpenAI argues that since powerful technology will inevitably be used by bad actors, there is a collective responsibility to act proactively. The company suggests that the flaws might not have been exploitable without AI, but the Defender's Window framework prioritizes getting ahead of the curve rather than waiting for an attack to occur.

The stakes for web developers and site administrators are high, as every site using PHP must upgrade to the latest patched version to remain secure. The incident serves as a case study for the broader impact of AI on cybersecurity. While some may argue that flaws which require AI to exploit are less dangerous, OpenAI maintains that the proactive approach is essential. The company points to similar recent findings by other AI models, such as Anthropic's Claude Mythos, which identified thousands of zero-day vulnerabilities in major operating systems and browsers. This trend suggests that AI-assisted security audits are becoming a standard and necessary part of maintaining robust digital infrastructure.

For users and developers in the Midwest and beyond, the immediate action required is to update PHP installations to the latest versions released by the Foundation. The rapid patching process demonstrates that the security community can respond quickly when AI tools are leveraged effectively. However, it also highlights the pressure on open-source projects to keep pace with the accelerating speed of AI-driven discovery. As these tools become more common, the gap between vulnerability discovery and exploitation continues to shrink, making proactive defense a critical priority for any organization relying on web technologies.