70% of the Web: AI Finds Flaw That Stumped Humans for Two Years

The PHP Foundation released emergency updates for four versions of its core software on October 21, 2026, bringing the release forward from its scheduled date of October 22. This accelerated timeline was necessitated by premature disclosures that occurred on October 20, which triggered public concern about the security of a programming language estimated to power 70% of websites. The patches, designated as PHP 8.2.35, 8.3.36, 8.4.27, and 8.5.12, address two critical vulnerabilities that researchers argue have been present since 2024.

The discovery was made under OpenAI’s “Patch the Planet” cybersecurity initiative, which utilizes the company’s GPT-5.6 Sol model to identify security gaps. The process began on July 30, 2026, when researchers from Calif, a California-based cybersecurity firm and founding partner of the program, started auditing PHP at the request of the PHP Foundation. Although the Foundation had been aware of an underlying flaw since 2024, human developers had been unable to determine a method for exploitation or a viable fix for the past two years.

On August 11, 2026, Calif researcher Ryan A. identified the first exploitable vulnerability, with OpenAI researchers confirming the finding the same day. Shortly thereafter, another cybersecurity firm, Wiz, independently identified a second, related vulnerability. The two issues, tracked as CVE-2026-109618 and CVE-2026-109619, stem from errors in how PHP manages memory and processes fatal errors. Both flaws received a critical severity score of 9.2 out of 10, indicating a high risk of remote exploitation without the need for login credentials.

The first vulnerability occurs when PHP exhausts its allocated memory during a request, potentially leaving behind incomplete data that the engine later misinterprets as valid. The second involves a fatal error leaving the system pointing to memory that has already been freed, allowing subsequent requests to access or corrupt that space. Because these flaws reside in the core engine, they can affect websites regardless of the security quality of their specific application code. Attackers can leverage these weaknesses to execute malicious code, potentially taking control of servers and exposing sensitive data such as passwords, authentication tokens, and API keys stored in memory.

OpenAI describes this incident as a demonstration of its “Defender's Window” concept. Introduced after its models hacked Hugging Face, the concept posits a narrow timeframe where organizations can use advanced AI to find and patch defenses before malicious actors gain access to similar capabilities. OpenAI argues that while human teams struggled for two years to resolve the code, AI-assisted researchers developed working exploits in just two weeks. The company emphasizes that proactive action is a collective responsibility, noting that Anthropic’s Claude Mythos recently identified thousands of zero-day vulnerabilities in major operating systems and web browsers.

Following the discovery, OpenAI coordinated with PHP maintainers, security providers, and government cybersecurity agencies to develop fixes over a two-month period. The company attributed the length of this timeline to the complexity of the task, which required significant changes to the PHP engine. While PHP versions 8.1 and older will not receive official fixes, OpenAI states that all sites using the language can now safely upgrade to the newly released versions to secure their infrastructure.

The incident has sparked debate regarding the role of AI in cybersecurity. Some argue that the flaws would not have been exploitable without AI, rendering them effectively harmless in a pre-AI context. OpenAI challenges this view, asserting that revolutionary technology will inevitably be used by bad actors. The company highlights the discovery as one of the most significant cybersecurity events of 2026, given the widespread ubiquity of PHP across the global web.