AI Tools Turn Two-Year-Old PHP Bug into Immediate Threat for Most of the Web

The digital backbone of the internet just got a major security overhaul. On October 21, the PHP Foundation released urgent updates for the programming language that powers an estimated 70% of all websites. This move was triggered by the discovery of two critical vulnerabilities that had sat unnoticed for two years. The breakthrough came from a coalition of researchers who used artificial intelligence to transform a long-standing code issue into a working exploit, prompting a coordinated fix that brought the release date forward by one day due to early public disclosures.

The latest development involves the specific versions of PHP that need attention. The Foundation has issued patched releases for versions 8.2.35, 8.3.36, 8.4.27, and 8.5.12. Users of PHP 8.1 and older versions will not receive official fixes, meaning they must upgrade to newer versions to stay secure. OpenAI stated that any site using PHP can now safely upgrade to protect itself against these specific threats.

To understand why this is significant, it helps to look back at how the flaw was found. The underlying bug in PHP’s code has been known to the Foundation since 2024. However, human developers and security teams struggled for two years to find a way to exploit it or to craft a proper fix. In late July 2026, the Foundation requested that Calif, a cybersecurity firm and founding partner of OpenAI’s “Patch the Planet” initiative, begin auditing the language using AI tools. Within weeks, the AI-assisted researchers identified the exploitable vulnerabilities on August 11, a task that had stumped human experts for months.

The two flaws, tracked as CVE-2026-109618 and CVE-2026-109619, relate to how PHP handles memory and errors. The first issue occurs when the engine runs out of allocated memory during a request, potentially leaving behind incomplete data that it later mistakes for valid information. The second 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 space. Both vulnerabilities carry 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. This could allow them to execute malicious code on affected servers and potentially take control of websites. The AI researchers also demonstrated that these 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 of PHP, even websites with perfectly secure application code are at risk.

The discovery highlights a concept OpenAI calls the “Defender’s Window.” This idea, introduced after the company’s models previously hacked the platform Hugging Face, describes a narrow timeframe where organizations can use advanced AI to find and patch defenses before bad actors can. OpenAI argues that while powerful technology will inevitably be used by attackers, there is a collective responsibility to act proactively. The company views this incident as one of the biggest cybersecurity discoveries of 2026, given how widely PHP is used.

The timeline for the fix was complex. After the initial discovery in August, OpenAI coordinated a group of PHP maintainers, security providers, and government agencies. The process took two months because the fixes required major changes to the PHP engine. The coordinated release was originally planned for October 22, but premature disclosures on October 20 caused public outcry, leading the team to move the release up to October 21.

Some might argue that since AI was needed to make these flaws exploitable, they were effectively harmless before now. However, OpenAI challenges this view. It points to recent findings by Anthropic’s Claude Mythos, which identified thousands of zero-day vulnerabilities in major operating systems and web browsers. The prevailing view among these tech leaders is that if a flaw can be exploited with current AI, it will eventually be exploited, making proactive patching essential.

For website owners and developers, the immediate action is clear: check which version of PHP is being used. If it is one of the patched versions listed above, an update is necessary. This event underscores the changing landscape of cybersecurity, where AI tools are no longer just for building software but are becoming essential for securing it against threats that human developers alone may miss.