OpenAI claims AI model found critical PHP bugs that humans missed for two years

OpenAI says its artificial intelligence model just solved a cybersecurity puzzle that stumped human experts for two years.

The announcement, made on October 21, details how researchers under the company's "Patch the Planet" initiative used GPT-5.6 Sol to discover critical vulnerabilities in PHP. This open-source language powers an estimated 70% of the internet. The discovery triggered a rapid coordination effort involving PHP maintainers, security providers, and government agencies to develop fixes over a two-month period. OpenAI attributes the timeline to the complexity of the task, which required significant changes to the core PHP engine.

The timeline of this discovery reveals a persistent gap in human analysis. The PHP Foundation first identified the underlying flaw in 2024 but could not determine how to exploit it or how to fix it. Human developers examined the code repeatedly but failed to resolve the issue. In July 2026, at the Foundation's request, researchers from Calif, a California-based cybersecurity firm and founding partner of the Patch the Planet program, began auditing PHP with OpenAI tools. On August 11, a Calif researcher identified the exploitable vulnerability. OpenAI researchers confirmed this finding the same day. Shortly after, another firm, Wiz, independently identified an additional related vulnerability.

The two flaws are tracked as CVE-2026-109618 and CVE-2026-109619. Both involve errors in how PHP manages memory and handles errors. The first occurs when the engine runs out of allocated memory during a request, potentially leaving incomplete data that the system later treats as valid. The second involves a fatal error that leaves PHP pointing to memory that has already been freed. This allows subsequent requests to access or corrupt that memory. Under specific conditions, attackers can exploit these weaknesses remotely without login credentials to execute malicious code and take control of websites. Both vulnerabilities received a critical severity score of 9.2 out of 10.

The implications extend beyond simple code errors. Because these flaws exist in the core PHP engine, websites can be compromised even if their application code is secure. Using AI, researchers demonstrated that the bugs could expose sensitive data stored in server memory, including passwords, authentication tokens, and API keys. The PHP Foundation has released patched versions 8.2.35, 8.3.36, 8.4.27, and 8.5.12 to address the issues. Versions 8.1 and older will not receive official fixes. The coordinated release was originally scheduled for October 22 but was brought forward to October 21 after premature disclosures on October 20 sparked public outcry.

OpenAI frames this incident as a prime example of its "Defender's Window" concept. The company introduced this idea after its models hacked Hugging Face. The concept describes a narrow timeframe where organizations can use advanced AI to find and patch defenses before malicious actors gain similar capabilities. OpenAI argues that while some might view the flaws as harmless since they would not be exploitable without AI, the company believes it is a collective responsibility to act proactively. The firm notes that powerful technology will inevitably be used by bad actors, making early detection crucial.

This incident highlights the shifting dynamics in cybersecurity. Human teams identified the bug in 2024 but could not weaponize it. AI-assisted researchers turned that known flaw into a working exploit in just two weeks. OpenAI cites this speed as evidence that AI tools are changing the pace of vulnerability discovery. The company positions PHP, given its ubiquity, as one of the most significant cybersecurity discoveries of 2026. Other AI models are showing similar capabilities. Anthropic's Claude Mythos recently identified thousands of zero-day vulnerabilities across major operating systems and web browsers.

For New England businesses and developers, the message is clear: update now. OpenAI states that every site using PHP can safely upgrade its version to stay secure. The incident serves as a reminder that legacy code and standard languages are under new scrutiny. As AI models become more sophisticated in finding and exploiting code flaws, the window for defenders to act is shrinking. Organizations relying on PHP must ensure they are running the latest patched versions to mitigate the risk of remote code execution and data exposure.

The rapid response from the PHP Foundation and its partners underscores the growing reliance on AI in the security sector. What took humans two years to identify as a problem took AI two weeks to prove was an exploit. This shift in capability is likely to accelerate the pace of future security audits. For now, the immediate priority is clear: patch, upgrade, and verify that no sensitive data was exposed during the period when these vulnerabilities were present.