OpenAI Claims AI Uncovered Critical PHP Flaws Affecting Majority of Web

A security vulnerability that stumped human developers for two years was transformed into a working exploit in just two weeks, according to OpenAI, which credits its artificial intelligence model for the breakthrough. The discovery highlights a growing tension in cybersecurity: while AI tools can uncover deep-seated flaws that humans miss, they also accelerate the timeline for attackers to leverage those same weaknesses. OpenAI announced on Wednesday that its GPT-5.6 Sol model identified critical flaws in PHP, the open-source programming language that underpins an estimated 70% of the world’s websites.

The incident centers on two distinct memory management errors in the PHP core engine, tracked as CVE-2026-109618 and CVE-2026-109619. Both vulnerabilities allow remote attackers to execute malicious code on affected servers without requiring login credentials. The first flaw occurs when PHP exhausts its allocated memory during a request, potentially leaving behind incomplete internal data that the engine later misinterprets as valid. The second involves a fatal error that leaves the system pointing to memory that has already been freed, enabling subsequent requests to access or corrupt that space. Both issues were assigned a critical severity score of 9.2 out of 10.

The discovery was made by researchers from Calif, a California-based cybersecurity firm and founding partner in OpenAI’s “Patch the Planet” initiative, who began auditing PHP on July 30 at the request of the PHP Foundation. The Foundation had been aware of the underlying issue since 2024 but lacked a method to exploit it or a viable fix. Calif researcher Ryan A. identified the exploitable vulnerability on August 11, a finding that was independently confirmed by OpenAI researchers the same day. Shortly after, another security firm, Wiz, identified an additional related vulnerability, completing the pair of critical flaws.

Following the identification of the bugs, OpenAI coordinated a coalition of PHP maintainers, security providers, and government cybersecurity agencies to develop and deploy fixes. The PHP Foundation released patched versions, including 8.2.35, 8.3.36, 8.4.27, and 8.5.12, on Tuesday. Versions 8.1 and older will not receive official patches. The coordinated release was originally scheduled for Wednesday but was moved up to Tuesday after premature disclosures on Monday sparked public outcry. OpenAI stated that every site using PHP can now safely upgrade to secure their systems.

OpenAI framed this incident as a prime example of its “Defender’s Window” concept, a term the company introduced after its models were used to hack Hugging Face. The concept describes a narrow timeframe during which organizations can use frontier AI models to find and patch defenses before malicious actors gain access to similar capabilities. The company argues that powerful technology will inevitably be used by bad actors, making it a collective responsibility to act proactively. This stance follows recent reports that Anthropic’s Claude Mythos identified thousands of zero-day vulnerabilities in major operating systems and web browsers.

The rapid exploitation of these flaws contrasts sharply with the two-year period during which human teams examined the code without a full resolution. While some argue that the flaws were effectively harmless if they could only be exploited with AI assistance, OpenAI maintains that the ability to turn theoretical bugs into practical attacks quickly changes the risk calculus. The company views the PHP incident as one of the most significant cybersecurity discoveries of 2026, given the language’s ubiquity across the global web infrastructure.