AI Finds Major Web Security Flaws
New AI tools helped researchers turn a long-standing code issue into a working exploit, forcing a rapid security update for most websites.
By Aminata Jenkins ·

If you run a website, the underlying software you rely on has just undergone a significant security overhaul. OpenAI announced on Oct. 21 that its advanced AI model, GPT-5.6 Sol, played a central role in discovering critical vulnerabilities in PHP, the programming language that powers an estimated 70% of the internet. For ordinary users, this means the sites they visit daily are likely more secure now, provided their hosts have updated their systems. The discovery highlights a shifting landscape where artificial intelligence is not just writing code but actively hunting for weaknesses that human teams have missed for years.
The core of this development is the speed with which the AI transformed a theoretical flaw into a practical threat. Human developers had known about the underlying issue since 2024 but were unable to figure out how to exploit it or fix it. When researchers from the cybersecurity firm Calif began using OpenAI’s tools on July 30, they identified an exploitable vulnerability just over a month later, on Aug. 11. Within two weeks of that initial find, the team had demonstrated working exploits, a process that had stumped human experts for two years.
The specific bugs, tracked as CVE-2026-109618 and CVE-2026-109619, are deeply embedded in how PHP handles memory and errors. The first issue occurs when the system runs out of allocated memory, potentially leaving behind incomplete data that the engine later mistakes for valid information. The second involves a fatal error that leaves the system pointing to memory that has already been freed, allowing subsequent requests to access or corrupt that space. Because these flaws exist in the core engine, a website can be vulnerable even if its own application code is perfectly secure.
The severity of these bugs is high, with both receiving a critical severity score of 9.2 out of 10. Attackers can exploit these weaknesses remotely without needing login credentials, potentially executing malicious code to take control of servers. Furthermore, the researchers demonstrated that these flaws could expose sensitive data stored in server memory, such as passwords, authentication tokens, and API keys. This makes the vulnerabilities particularly dangerous for businesses and individuals whose data is hosted on PHP-based platforms.
In response to the discovery, the PHP Foundation released patched versions 8.2.35, 8.3.36, 8.4.27, and 8.5.12. It is important to note that PHP 8.1 and older versions will not receive official fixes, meaning users on those systems must upgrade to stay secure. The coordinated release was originally scheduled for Oct. 22 but was moved up to Oct. 21 after premature disclosures on Oct. 20 caused public concern. OpenAI emphasized that every site can now safely upgrade to protect itself.
OpenAI framed this incident as a prime example of what it calls the "Defender's Window." This concept describes a narrow timeframe where organizations can use frontier AI models to find and patch weaknesses before those same AI-driven attack capabilities become widely available to malicious actors. The company argues that powerful technology will inevitably be used by bad actors, so it is a collective responsibility to act proactively. This perspective challenges the idea that flaws not yet exploited by humans are harmless, suggesting instead that AI effectively makes them exploitable.
The timeline of the fix was lengthy, taking two months to coordinate among researchers, maintainers, security providers, and government agencies. OpenAI attributed this duration to the complexity of the task, which required major changes to the PHP engine. The collaboration involved the founding partner of the "Patch the Planet" program, Calif, and the cybersecurity firm Wiz, which independently identified an additional vulnerability. This multi-party effort underscores the growing complexity of modern cybersecurity, where AI tools and human expertise must work in tandem to secure the digital infrastructure.
Looking ahead, this event signals a broader trend in the tech industry. With other AI models, such as Anthropic's Claude Mythos, recently identifying thousands of zero-day vulnerabilities in major operating systems and browsers, the pace of discovery is accelerating. For the Southwest and the broader US market, this means a heightened focus on software maintenance and upgrade cycles. As AI becomes more integrated into security workflows, the ability to quickly adapt to these new findings will be crucial for maintaining trust and safety in the digital world.






