AivexaNewsSearch
AI news for builders and product teamsChecked every hour

OpenAI's GPT-5.6 Sol and Google's Gemini 4 Argon found critical PHP flaw affecting millions of websites

Collected Oct 21, 2026

OpenAI's GPT-5.6 Sol discovered a critical PHP vulnerability as part of OpenAI's Patch-the-Planet initiative, which the company describes as one of the most impactful security issues in recent memory. Security firm Wiz independently found the same flaws, working with its own Wiz Atlas platform alongside Google's Gemini 4 Argon and GPT-5.6 Sol. The vulnerability could affect millions of websites worldwide.

Because the models surfaced the flaws before attackers did, developers and security vendors gained time to respond. A tool published on GitHub now lets server operators check whether their PHP version is affected.

According to Wiz, Gemini 4 Argon proved especially useful across several stages of the work: finding the vulnerability, developing exploits, building attack detection, and validating the patch. Two models arriving at the same flaw independently is notable, because it suggests the finding is unlikely to be an artifact of one model's behavior.

The remediation was coordinated over roughly two months across five phases. Participants included PHP developers, Wiz, Cloudflare, and US agencies including CISA and the White House. OpenAI is proposing the process as a repeatable framework for future security discoveries driven by AI.

Wiz's analysis points to a shift in the economics of exploitation. Adapting an exploit that worked against one specific PHP installation to other versions and configurations used to take significant effort. Large language models remove that barrier, Wiz says, because they can automatically adjust exploits across different PHP versions, configurations, and memory layouts. The firm writes that this significantly lowers the barrier to scaling attacks across deployments, and that vulnerabilities once considered impractical to exploit become viable targets.

OpenAI expects findings on the scale of the PHP2Shell flaw to increase sharply in the coming months. The company also warns that the same capabilities that uncover vulnerabilities can help attackers. That tension sits inside a months-long debate over how well AI models perform on offensive versus defensive cybersecurity work.

For context, PHP remains one of the most widely deployed server-side languages powering the web, which is why a flaw in its handling of certain inputs can ripple across a very large installed base. Coordinated disclosure—where researchers and vendors agree on a timeline before public details appear—is the standard mechanism for limiting exposure in cases like this. What stands out here is that the discovery phase was itself partly automated, while the disclosure and patching phase still depended on human coordination among vendors, maintainers and government agencies.

This suggests a likely trade-off for defenders. Model-assisted discovery can compress the time between a flaw existing and a patch shipping, but the same models lower the cost of weaponizing what they find. In practice, teams running PHP in production should treat the GitHub checker as an immediate action item, then apply the upstream fix and review whether their deployment configurations differ from the patched baseline—Wiz's point about configurations is that a patch validated in one setup may need revalidation in another.

Why it matters: Server operators running PHP, and the platform and security teams that support them, face a near-term patching task and a longer-term change in how vulnerabilities are found. For developers, the practical signal is that AI-assisted disclosure may shorten the window between discovery and coordinated fix, while also raising the odds that exploits adapt quickly across versions—an inference from Wiz's stated concerns rather than a reported outcome. Product teams should expect such framework-style processes, involving vendors and agencies, to become more common as model-driven security work scales.

Read at The Decoder

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

OpenAI's GPT-5.6 Sol discovered a critical PHP vulnerability as part of the Patch the Planet initiative that could affect millions of websites. Security firm Wiz independently found the same flaw using Google's Gemini 4 Argon. OpenAI warns that the same AI capabilities that find vulnerabilities can help attackers too. The article OpenAI's GPT-5.6 Sol and Google's Gemini 4 Argon found critical PHP flaw affecting millions of websites appeared first on The Decoder .