OpenAI Leverages AI to Expose Critical PHP Flaws Affecting Most of the Web

OpenAI announced on October 21 that its AI model, GPT-5.6 Sol, played a central role in identifying critical security vulnerabilities in PHP. The language, which is estimated to power 70% of the internet, contained two significant flaws that could allow attackers to take control of servers. The discovery was part of the company's Patch the Planet cybersecurity initiative, which aims to use advanced AI tools to find and fix weaknesses before they are widely exploited by malicious actors.

The main claim from OpenAI is that the AI system successfully converted a bug that had stumped human developers for two years into a practical exploit. Researchers began auditing the PHP codebase on July 30, 2026, at the request of the PHP Foundation. By August 11, a researcher from the California-based firm Calif, a founding partner of the initiative, identified the first exploitable vulnerability. OpenAI researchers confirmed the finding the same day, and the cybersecurity firm Wiz independently discovered a second related flaw. These two issues, tracked as CVE-2026-109618 and CVE-2026-109619, both involve errors in how the PHP engine manages memory and handles errors.

The vulnerabilities are severe because they reside in the core engine of PHP. This means that a website can be vulnerable even if its specific application code is secure. One flaw occurs when PHP runs out of allocated memory, potentially leaving behind incomplete data that the engine later treats as valid. The other involves a fatal error that leaves the system pointing to memory that has already been freed. Attackers can exploit these weaknesses remotely without login credentials to execute malicious code. Both issues received a critical severity score of 9.2 out of 10, and researchers demonstrated that they could expose sensitive data such as passwords and API keys stored in server memory.

The PHP Foundation has released patched versions 8.2.35, 8.3.36, 8.4.27, and 8.5.12 to address the problems. Versions 8.1 and older will not receive official fixes. The coordinated release was originally planned for October 22, but it was moved up to October 21 after premature disclosures on October 20 caused public concern. OpenAI stated that the two-month timeline for fixing the issues was due to the complexity of the task, which required major changes to the PHP engine. The company coordinated with a coalition of researchers, maintainers, security providers, and government agencies to ensure a smooth rollout.

OpenAI frames this event as a demonstration of its Defender's Window concept. This idea suggests there is a narrow timeframe during which organizations can use frontier AI models to strengthen their defenses before those same capabilities become available to threat actors. The company argues that powerful technology will inevitably be used by bad actors, so it is a collective responsibility to act proactively. It cites a recent discovery by Anthropic's Claude Mythos, which identified thousands of zero-day vulnerabilities in major operating systems and web browsers, to support this view.

There are caveats to the narrative. Human developers had examined the underlying code since 2024 but could not fully resolve the flaws. Some might argue that the bugs would not have been exploitable without AI, making them effectively harmless in the traditional sense. However, OpenAI challenges this view by emphasizing that the ability to exploit the flaws is now a reality. The company considers this one of the biggest cybersecurity discoveries of 2026 due to the ubiquity of PHP. All sites using the language are advised to upgrade to the latest patched versions to remain secure.

The incident highlights the shifting dynamics of cybersecurity in the age of advanced AI. The speed at which the researchers moved from identifying the flaw to demonstrating a working exploit, taking only two weeks, contrasts sharply with the two-year period during which the bug existed without a known fix. This rapid cycle of discovery and exploitation underscores the urgency of using AI tools defensively. As AI models become more capable, the gap between finding a vulnerability and exploiting it may continue to shrink, putting pressure on developers and security teams to adapt their workflows.

For the broader tech industry, this event serves as a wake-up call regarding the reliance on legacy code and the potential of AI to uncover hidden risks. The cooperation between OpenAI, the PHP Foundation, and private cybersecurity firms sets a precedent for how such discoveries might be handled in the future. By bringing the fix forward in response to public outcry, the stakeholders demonstrated a commitment to transparency and rapid action. As more organizations adopt AI for security audits, similar discoveries may become more common, requiring ongoing vigilance and quick response strategies from all parties involved.