OpenAI reveals AI found major PHP security flaws

OpenAI announced on October 21 that its GPT-5.6 Sol model played a pivotal role in identifying critical security vulnerabilities within the PHP programming language. The discovery was part of the company's "Patch the Planet" cybersecurity initiative, which aims to leverage artificial intelligence to strengthen digital defenses. According to the company, researchers used the AI to uncover flaws that had eluded human developers for two years, ultimately turning a known bug into a working exploit within two weeks.

The vulnerabilities were discovered by researchers from Calif, a California-based cybersecurity firm and the founding partner of the Patch the Planet program. At the request of the PHP Foundation, the team began auditing the software on July 30, 2026. On August 11, Calif researcher Ryan A. identified the exploitable nature of the bug, a finding that was corroborated the same day by OpenAI’s internal researchers. Shortly after, the cybersecurity firm Wiz independently identified a related additional vulnerability. Both flaws are tracked as CVE-2026-109618 and CVE-2026-109619 and carry a critical severity score of 9.2 out of 10.

PHP is an open-source language that powers an estimated 70% of the internet, making these discoveries particularly significant. The flaws reside in the core engine’s memory management and error handling processes. One issue arises when PHP exhausts its allocated memory during request processing, potentially leaving incomplete internal data that the engine later misinterprets as valid. The second involves a fatal error causing the engine to point to memory that has already been freed, allowing subsequent requests to access or corrupt that data. Because these weaknesses exist in the core engine, websites can be compromised even if their specific application code is secure.

The PHP Foundation has released patched versions, including 8.2.35, 8.3.36, 8.4.27, and 8.5.12, to address these issues. Versions 8.1 and older will not receive official fixes. While the coordinated release was initially scheduled for October 22, it was moved up to October 21 due to premature disclosures on October 20 that triggered public concern. OpenAI stated that the two-month timeline required for the fix was necessary because the task involved major changes to the PHP engine, a complexity that had previously stumped human security teams.

OpenAI framed this incident as a prime example of its "Defender's Window" concept. Introduced after its models hacked Hugging Face, the concept describes a narrow timeframe during which organizations can use advanced frontier AI to find and patch vulnerabilities before those same capabilities become widely available to malicious actors. The company argues that since powerful technology will inevitably be used by bad actors, there is a collective responsibility to act proactively. This perspective contrasts with the view that flaws unexploitable without AI are effectively harmless, a stance OpenAI challenges by citing recent findings by Anthropic’s Claude Mythos regarding zero-day vulnerabilities in major operating systems and browsers.

The researchers demonstrated that the flaws could be exploited remotely without login credentials, allowing attackers to execute malicious code and potentially take control of affected servers. Furthermore, the vulnerabilities could expose sensitive information stored in server memory, such as passwords, authentication tokens, and API keys. OpenAI highlighted the incident as one of the most significant cybersecurity discoveries of 2026, given the ubiquity of PHP.

Moving forward, OpenAI advises that every site using PHP can safely upgrade to the latest patched versions to maintain security. The company continues to emphasize the importance of the Defender's Window, urging organizations to utilize advanced AI models to strengthen their defenses while the technology remains a strategic advantage in the cybersecurity landscape.