AI Models Expose Critical Flaws in PHP Code Used by Most of the Web

The internet runs on PHP, a foundational piece of software that supports an estimated seventy percent of all websites. On October 21, OpenAI announced that its artificial intelligence model, GPT-5.6 Sol, played a central role in uncovering two severe security vulnerabilities within that codebase.

The discovery was made by researchers participating in the Patch the Planet cybersecurity initiative. These specialists used the AI tool to identify weaknesses in how PHP manages computer memory and processes errors. The findings were significant enough to trigger a coordinated response from PHP maintainers, security providers, and government agencies. The goal was to develop and deploy fixes before the flaws could be widely exploited.

The vulnerabilities, designated as CVE-2026-109618 and CVE-2026-109619, stem from fundamental issues in the PHP engine. The first occurs when the system runs out of allocated memory during a request. This can leave behind incomplete internal data that the engine later mistakes for valid information. The second issue arises when a fatal error causes PHP to point to memory that has already been freed. This allows subsequent requests to access or corrupt that data. Both flaws received a critical severity score of 9.2 out of 10.

These bugs are particularly dangerous because they reside in the core engine of the language. This means a website can be vulnerable even if its specific application code is perfectly secure. Attackers could potentially exploit these weaknesses remotely without needing login credentials. Such an attack could allow them to execute malicious code on affected servers. The researchers also demonstrated that the flaws could expose sensitive data stored in server memory, including passwords, authentication tokens, and API keys.

The PHP Foundation responded by releasing patched versions for PHP 8.2.35, 8.3.36, 8.4.27, and 8.5.12. Older versions, including PHP 8.1 and earlier, will not receive official fixes. The release was originally scheduled for October 22. However, it was moved up to October 21 after premature disclosures on October 20 led to public concern. OpenAI stated that the lengthy two-month timeline for the fix was necessary due to the complexity of the required changes to the PHP engine.

The underlying flaw had actually been known to the PHP Foundation since 2024. Human developers and security teams had examined the code but could not identify a method to exploit it or create a fix. The situation changed when researchers from Calif, a California-based cybersecurity firm and founding partner of the Patch the Planet program, began auditing PHP with OpenAI tools on July 30. They started this work at the request of the Foundation.

On August 11, Calif researcher Ryan A. identified an exploitable vulnerability. OpenAI researchers confirmed this finding on the same day. Shortly after, another cybersecurity firm, Wiz, independently identified an additional vulnerability. The AI-assisted team demonstrated practical ways to exploit the flaws within two weeks. This stood in contrast to the two years the issue had persisted without a solution.

OpenAI frames this incident as a prime example of its concept known as the Defender's Window. The company describes this as a narrow timeframe during which organizations can use advanced AI models to find and patch weaknesses before malicious actors gain access to the same capabilities. OpenAI argues that it is a collective responsibility to act proactively. The company notes that Anthropic's Claude Mythos recently identified thousands of zero-day vulnerabilities in major operating systems and web browsers, highlighting the growing scope of AI-driven security scanning.

The company maintains that powerful technology will inevitably be used by bad actors. Therefore, finding and fixing these flaws now is essential. OpenAI considers this one of the biggest cybersecurity discoveries of 2026, given the ubiquity of PHP. The incident underscores the shifting dynamics of software security, where AI tools are rapidly changing the landscape for both defenders and attackers.

For now, every site using PHP can safely upgrade its version to stay secure. The rapid coordination between industry partners and government bodies demonstrates a new model for addressing critical software flaws. As AI models become more sophisticated, the pace of vulnerability discovery and patching is likely to accelerate further. This event may serve as a turning point in how the industry approaches the security of foundational web technologies.