AI models exposed a hidden PHP flaw that stumped human experts for two years

An estimated 70% of all websites on the internet were potentially exposed to a critical security flaw that human developers could not fix for two years. OpenAI announced on Oct. 21 that its artificial intelligence model, GPT-5.6 Sol, identified the vulnerability within the core of PHP, the open-source programming language that underpins a massive portion of the web. This discovery has triggered a coordinated global effort to patch the issue, which security experts are calling one of the most significant cybersecurity events of 2026.

The central claim from OpenAI is that advanced AI can now detect and exploit complex software weaknesses that have eluded human teams for long periods. The company stated that researchers working under its "Patch the Planet" initiative found the flaw in July 2026. While the PHP Foundation had been aware of the underlying issue since 2024, they were unable to determine how to exploit it or how to correct it. It was not until Aug. 11, 2026, that researchers from Calif, a founding partner of the program, identified the exploitable version of the bug with the help of the AI tool.

The vulnerabilities, designated as CVE-2026-109618 and CVE-2026-109619, stem from how PHP manages computer memory and handles errors. The first flaw occurs when the system runs out of allocated memory during a request, potentially leaving behind incomplete data that the engine later misinterprets as valid information. The second issue involves a fatal error that causes PHP to point to memory that has already been freed. This allows subsequent requests to access or corrupt that data. Both bugs received a critical severity score of 9.2 out of 10.

These are not minor glitches. Attackers can exploit these weaknesses remotely without needing login credentials to execute malicious code on affected servers. This could allow them to take control of websites entirely. The researchers also demonstrated that the flaws could expose sensitive data stored in server memory, such as passwords, authentication tokens, and API keys. Because the bugs exist in the core engine, websites can be affected even if their specific application code is written perfectly.

The PHP Foundation released patched versions 8.2.35, 8.3.36, 8.4.27, and 8.5.12 to address the issues. Versions 8.1 and older will not receive official fixes, meaning sites running those older versions must upgrade to remain secure. The coordinated release was originally scheduled for Oct. 22, but it was moved up to Oct. 21 after premature disclosures on Oct. 20 caused public concern. OpenAI says every site using PHP can now safely upgrade to stay protected.

OpenAI attributes this incident to what it calls the "Defender's Window." This concept describes a narrow timeframe where organizations can use advanced AI models to find and patch security defenses before those same capabilities become widely available to malicious actors. The company argues that powerful technology will inevitably be used by bad actors, so it is a collective responsibility to act proactively. OpenAI noted that Anthropic’s Claude Mythos recently identified thousands of zero-day vulnerabilities in major operating systems and web browsers, supporting the idea that AI is rapidly changing the security landscape.

The timeline of the fix was extensive, taking two months to coordinate. This was due to the complexity of the task, which required major changes to the PHP engine. A coalition of researchers, maintainers, security providers, and government cybersecurity agencies worked together to resolve the issue. The company highlighted that while human developers examined the code for two years, AI-wielding researchers demonstrated practical exploitation methods in just two weeks.

There are caveats to this narrative. One could argue that the flaws would never have been exploitable without the assistance of AI, which might make them effectively harmless in a pre-AI context. However, OpenAI challenges this view by emphasizing the speed at which AI can turn theoretical bugs into working exploits. The incident underscores a shifting dynamic in cybersecurity, where the gap between finding a bug and exploiting it is closing rapidly.

For the average user or site owner, the immediate takeaway is the importance of keeping software up to date. With the patches now available, the risk of these specific vulnerabilities is mitigated for those who update. However, the incident serves as a broader warning that the tools used to secure our digital infrastructure are evolving just as fast as the tools used to break it.

This event marks a significant milestone in the integration of AI into cybersecurity practices. It demonstrates both the power and the potential risks of relying on advanced models to maintain the integrity of the internet's foundational software. As AI capabilities continue to grow, the pressure on developers and security teams to keep pace will likely intensify.