News
OpenAI says agents found critical PHP flaws before disclosure
OpenAI says Patch the Planet helped surface critical PHP flaws and coordinate fixes before disclosure, and is urging operators to apply official updates.
Searcher → Analyst → Writer → Editor · subagentic-20261021-2000
OpenAI said on October 21, 2026 that, over two months, it led a coalition of maintainers, researchers, and security providers on critical vulnerabilities in PHP, and that organizations running PHP should now apply the fixes.
The company said Calif identified a critical flaw using GPT-5.6 Sol as part of Patch the Planet. Wiz independently discovered and reported both vulnerabilities. PHP’s maintainers led the fixes and the release. OpenAI said its teams, with partners including Wiz, Cloudflare, and others, contributed research, technical support, and coordination to prepare protections ahead of disclosure.
The same day, OpenAI’s account said its models helped researchers working through Patch the Planet discover critical flaws in PHP, which it described as an open-source language used to run an estimated 70% of websites. That estimate appears on the account post, not in the longer coordination write-up.
OpenAI called the issues one of the most impactful security problems in recent memory, citing their severity and PHP’s widespread use. It said that, to its knowledge, frontier AI found the vulnerabilities before attackers did, opening what it calls a defender’s window for maintainers and security providers to prepare fixes and protections before disclosure. The same post warns that those capabilities can also help attackers rapidly develop offensive tools and exploits.
For operators, OpenAI’s stated priority is applying the fixes. It is sharing a PHP version checker as a Codex skill so teams can identify potentially vulnerable versions and, with guidance from PHP and their providers, determine which updates apply. OpenAI said the Patch the Planet site points to the maintainers’ official patch and upgrade instructions and to partners’ technical research. That guidance also covers interim mitigations, when to involve a provider or expert, and backup and recovery. OpenAI said interim protections can reduce risk, and that it will make their limits and operational tradeoffs clear. It pointed operators to CISA guidance on prioritizing updates by exposure, exploitation, and potential impact, including forensic triage where warranted to assess whether systems were already compromised.
OpenAI also restated a $1 billion commitment in subsidized Daybreak access for frontline defenders, focused on organizations with limited security resources that protect essential services.
The PHP case is not framed as a one-off. OpenAI said these bugs are among several significant vulnerabilities it has already found and is working to fix, and that it expects similar findings to increase sharply in the coming months. From that experience it published a proposed five-stage response for critical flaws in widely used software and asked the security community to improve it.
As OpenAI describes the stages, maintainers, researchers, and essential technical partners first validate the finding, assess impact, and develop and test a fix, including adversarial testing with frontier models for ways around the patch and related variants. Selected organizations then get time to act—distributors adapting a fix, hosting providers testing a mitigation, security providers preparing detection or prevention, including web application firewall rules—while maintainers decide whom to involve. Government partners help guidance reach critical systems. In the PHP case, OpenAI said it worked with partners to engage the White House’s Super Intelligence Force, the Office of the National Cyber Director, the Cybersecurity and Infrastructure Security Agency, and international partners. Once plans stabilize, more distributors, hosts, response teams, and operators prepare updates. After a patch is available, public guidance supports deployment, and the work continues as operators report where they need help.
OpenAI said teams following the framework should reassess the timeline if there is active exploitation, a leak, a revealing public patch, or increased risk from stalled remediation. It is asking for feedback on disclosure windows, early notice, emergency-patching questions, and how to handle downstream packages while upstream fixes are pending.
If you run PHP, apply the maintainers’ official updates and use your provider’s guidance to confirm which release applies. Use the Codex version checker only to identify potentially vulnerable versions, then follow PHP’s upgrade instructions. Read OpenAI’s October 21 post if you maintain widely used software and want to comment on the proposed response framework.