AI Models Expose Critical PHP Flaws Threatening Majority of Web

The internet’s most ubiquitous programming language just revealed a vulnerability that human experts could not solve for two years.

OpenAI disclosed on Wednesday that its GPT-5.6 Sol model, operating under the "Patch the Planet" cybersecurity initiative, discovered two critical flaws in PHP, the open-source language that powers an estimated 70% of the web. The discovery triggered a rapid, coordinated response among PHP maintainers, security providers, and government agencies. Patched versions, specifically PHP 8.2.35, 8.3.36, 8.4.27, and 8.5.12, were released on October 21, a day earlier than the originally scheduled October 22 date. This acceleration was necessary due to premature disclosures on October 20 that generated public concern. Users of PHP 8.1 and older versions are left without official fixes, a detail that raises questions about the longevity of the language's support cycle.

The vulnerabilities, tracked as CVE-2026-109618 and CVE-2026-109619, stem from fundamental errors in how PHP manages memory and handles errors. The first flaw occurs when the engine runs out of allocated memory, potentially leaving incomplete internal data that is later misinterpreted as valid. The second involves a fatal error causing PHP to point to memory that has already been freed, allowing subsequent requests to access or corrupt that space. Both issues carry a critical severity score of 9.2 out of 10. Because these bugs reside in the core engine rather than specific application code, websites can be compromised even if their own code is secure. Attackers can exploit these weaknesses remotely without login credentials to execute malicious code or expose sensitive data such as passwords, authentication tokens, and API keys.

The timeline of the discovery highlights the shifting dynamics of cybersecurity. Researchers from Calif, a California-based firm and founding partner of the Patch the Planet program, began auditing PHP with OpenAI tools on July 30, 2026, at the request of the PHP Foundation. The Foundation had been aware of the underlying flaw since 2024 but lacked a method to exploit or fix it. On August 11, Calif researcher Ryan A. identified an exploitable vulnerability, with OpenAI researchers following suit the same day. The cybersecurity firm Wiz independently discovered a second vulnerability during this period. While human teams had examined the code for two years without resolution, AI-assisted researchers developed working exploits within two weeks.

OpenAI frames this incident as a prime example of its "Defender's Window" concept, a theory introduced after its models hacked Hugging Face. The company argues that there is a narrow timeframe during which organizations can use frontier AI to fortify defenses before malicious actors gain similar capabilities. This perspective challenges the view that flaws which require AI to exploit are inherently harmless. OpenAI maintains that powerful technology will inevitably be used by bad actors, making proactive defense a collective responsibility. This stance follows recent reports that Anthropic's Claude Mythos identified thousands of zero-day vulnerabilities in major operating systems and web browsers.

The stakes for the web are significant. With PHP running the majority of the internet, any core engine vulnerability has a massive blast radius. The incident underscores a growing dependency on AI to uncover security holes that escape human detection, while simultaneously raising questions about the pace at which defenders can patch systems compared to the speed at which attackers can weaponize AI tools. As the patched versions roll out, the industry faces a critical test of whether the Defender's Window is wide enough to protect a landscape increasingly shaped by artificial intelligence.