OpenAI Claims AI Uncovered Two-Year-Old PHP Flaw Affecting 70% of the Web

Seventy percent of the world’s websites may be running on compromised code, a staggering figure that underscores the urgency of recent developments in internet security. OpenAI announced on Wednesday that its GPT-5.6 Sol model, utilized within the company’s “Patch the Planet” cybersecurity initiative, identified critical vulnerabilities in PHP, the open-source programming language that powers the majority of the web. The discovery highlights a shifting dynamic in cybersecurity, where artificial intelligence is not only detecting threats but actively bridging the gap between theoretical flaws and practical exploitation.

The technical details reveal two distinct but related memory management errors, tracked as CVE-2026-109618 and CVE-2026-109619. Both flaws reside in PHP’s core engine, meaning they affect sites even if the specific application code is secure. The first vulnerability occurs when the engine runs out of allocated memory during request processing, potentially leaving incomplete internal data that is later misinterpreted as valid. The second involves a fatal error causing PHP to point to memory that has already been freed, allowing subsequent requests to access or corrupt that space. Both issues were assigned a critical severity score of 9.2 out of 10, and researchers demonstrated that attackers could exploit them remotely without login credentials to execute malicious code or expose sensitive data like passwords and API keys.

The timeline of the discovery illustrates the rapid pace of AI-assisted security research. Researchers from Calif, a California-based cybersecurity firm and founding partner of the Patch the Planet program, began auditing PHP on July 30 at the request of the PHP Foundation. While the Foundation had been aware of the underlying flaw since 2024, human developers had failed to identify an exploitation method or a fix for two years. On August 11, Calif researcher Ryan A. identified the exploitable vulnerability, with OpenAI researchers confirming the finding the same day. A separate firm, Wiz, independently discovered an additional related vulnerability during this period. In contrast to the two-year human struggle, the AI-assisted team demonstrated practical exploitation methods within two weeks.

Following the discovery, OpenAI coordinated a coalition of PHP maintainers, security providers, and government cybersecurity agencies to address the issues over a two-month period. The complexity of the task required major changes to the PHP engine, which contributed to the lengthy timeline. The PHP Foundation released patched versions 8.2.35, 8.3.36, 8.4.27, and 8.5.12 to resolve the vulnerabilities. Notably, PHP 8.1 and older versions will not receive official fixes. The coordinated release was originally scheduled for today, October 22, but was brought forward to yesterday, October 21, after premature disclosures on October 20 triggered public outcry.

OpenAI positions 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 where organizations can use advanced frontier AI to strengthen defenses before those same capabilities become widely available to malicious actors. The company argues that it is a collective responsibility to act proactively, especially as other models like Anthropic’s Claude Mythos have recently identified thousands of zero-day vulnerabilities in major operating systems and browsers.

For New England businesses and digital agencies, the stakes are high. Since the vulnerabilities exist in the core engine, upgrading is the primary mitigation strategy. OpenAI advises that all sites using PHP can safely upgrade to the latest patched versions to remain secure. While some argue that flaws requiring AI to exploit are effectively harmless, OpenAI challenges this view, asserting that revolutionary technology will inevitably be used by bad actors, making proactive defense essential.