OpenAI leads coalition to fix critical PHP flaws found with GPT-5.6 Sol

Two PHP vulnerabilities, one coordinated response

According to OpenAI, Calif found a critical flaw with GPT-5.6 Sol under the Patch the Planet initiative, while Wiz independently discovered and reported both vulnerabilities. PHP's maintainers led the fixes and the release. OpenAI's teams and partners such as Wiz, Cloudflare and others contributed research, technical support and coordination so that protections were ready ahead of disclosure.

For organizations running PHP, OpenAI says the priority now is applying the fixes. The company is sharing a PHP version checker and plans to launch an accompanying Codex skill to help operators identify potentially vulnerable PHP versions and work out which updates apply, alongside guidance from PHP and their own providers.

OpenAI calls the issues "one of the most impactful security issues in recent memory", citing their severity and PHP's widespread use. It also says that, to its knowledge, frontier AI found these vulnerabilities before attackers did, which opened what it calls a Defender's Window for maintainers and security providers to prepare.

A five-stage framework for critical vulnerabilities

OpenAI is proposing an approach to responding to critical flaws in widely used software and is asking the security community to help improve it. The response can involve five stages:

• Validate and fix upstream. The maintainer, researchers and essential technical partners confirm the finding, assess impact, and develop and test a fix, including adversarial testing with frontier models to look for ways around the patch and related variants.
• Give affected organizations time to act. Maintainers decide whom to involve. A distributor might adapt a fix for an older release, a hosting provider might test a mitigation, and a security provider might prepare detection or prevention measures, including WAF rules.
• Coordinate with government partners. For PHP, OpenAI says it worked with partners to engage parts of the U.S. government, including the White House's Super Intelligence Force, ONCD and CISA, as well as international partners.
• Prepare updates across affected products. Once the fix and disclosure plans are stable, more distributors, hosting and service providers, response teams and operators can be brought in.
• Publish guidance and support deployment. Public guidance helps people assess exposure, find updates or interim mitigations, and verify that a fix is running.

The framework weighs an organization's protective role, the time it needs to act, and its ability to handle sensitive information. Participants receive what they need, from network logs with indicators of compromise to working proof-of-concept exploits. Because even limited details can reveal a vulnerability, OpenAI says it weighs the risk of sharing against the protection it enables. The timeline would be reassessed if there is active exploitation, a leak, a revealing public patch, or increased risk from stalled remediation.

The Patch the Planet website collects public disclosures and guidance. For PHP, it points to the maintainers' official patch and upgrade instructions and partners' technical research, and covers interim mitigations, backup and recovery. Operators can use CISA's guidance on prioritization and remediation to decide between emergency patching, immediate mitigation or a planned update.

Support for maintainers and defenders

OpenAI notes that many projects are run by small teams who handle security reports alongside routine development. It says it supports them with funding, access to capable models and hands-on engineering help. Its investments include subscriptions for all maintainers through Codex for OSS and direct sponsorship of projects such as the Rust Foundation. Open-source projects can apply to Patch the Planet for hands-on support.

The company has also committed $1 billion in subsidized Daybreak access through Daybreak for Frontline Defenders, aimed at organizations with limited security resources that protect essential services. OpenAI is calibrating additional blocking classifiers for Daybreak to reduce misuse, and acknowledges the tradeoff that safeguards against abuse can also block legitimate security work.

What it means for operators

OpenAI says these bugs are among several significant vulnerabilities it has already found and is working to fix, and it expects similar findings to "increase sharply in the coming months". For PHP users, the immediate step is to check versions and apply the maintainers' fixes. For the wider ecosystem, OpenAI argues that a repeatable process, backed by resources for maintainers, will be needed as AI-driven discovery scales, while warning that the same capabilities can help attackers build exploits quickly.