OpenAI's GPT-5.6 Sol Model Identifies Critical Flaws in PHP

For two years, a dangerous flaw sat dormant in PHP, the programming language that powers an estimated 70% of the internet's websites. While human developers and security teams were unable to devise a fix or a method of attack, a team of researchers armed with OpenAI's AI model identified not only how to exploit the vulnerability but also a second, related flaw, turning a theoretical problem into a working exploit in just two weeks.

The vulnerabilities, discovered under OpenAI's Patch the Planet cybersecurity program, were found in July and August 2026. Researchers from Calif, a California-based firm and founding partner of the initiative, began auditing PHP at the PHP Foundation's request. Calif researcher Ryan A. first identified an exploitable vulnerability on August 11, 2026, followed by OpenAI researchers the same day. Another security firm, Wiz, independently uncovered an additional vulnerability. The two flaws, tracked as CVE-2026-109618 and CVE-2026-109619, involve errors in how PHP manages computer memory.

The potential impact is severe. Both vulnerabilities received a critical severity score of 9.2 out of 10 and allow attackers to remotely execute malicious code on a server without needing login credentials, which could lead to website takeovers. Furthermore, the flaws are located in PHP's core engine, meaning a website is vulnerable even if its own application code is completely secure. Researchers also demonstrated that the vulnerabilities could expose sensitive data stored in server memory, such as passwords, authentication tokens, and API keys.

The PHP Foundation has released patched versions of the software, including PHP 8.2.35, 8.3.36, 8.4.27, and 8.5.12, to fix the issues. Official fixes will not be provided for PHP 8.1 and older versions. The coordinated security update was scheduled for release on October 22, 2026, but was moved up to October 21 after premature disclosures on October 20 led to public concern. OpenAI is urging every site that uses PHP to upgrade its version to remain secure.

This incident is being framed by OpenAI as a key example of what the company calls the Defender's Window. This concept, which OpenAI introduced after its models hacked Hugging Face, describes a critical and narrow timeframe where organizations can use advanced AI to find and patch security flaws before those same capabilities become widely available to malicious actors. The idea is that powerful technology will inevitably be used by bad actors, making it a collective responsibility to act proactively. OpenAI considers this one of the biggest cybersecurity discoveries of 2026, a stance it is reinforcing by pointing to other AI models, such as Anthropic's Claude Mythos, which recently identified thousands of zero-day vulnerabilities across major operating systems and web browsers.

The successful patching of these flaws highlights a growing divide in cybersecurity capabilities. While human experts struggled with the problem for two years, an AI-assisted approach resolved it in a matter of weeks. As AI models become more powerful, the ability to rapidly find and fix critical vulnerabilities before they can be weaponized is becoming an increasingly urgent priority for companies and security agencies.