Skip to main content
Back to timeline
OpenAISource publication:

OpenAI says frontier models found PHP flaws before attackers did, and proposes a five-stage disclosure playbook

Synopsis

Under its Patch the Planet initiative, OpenAI says GPT‑5.6 Sol was used to find a critical flaw in PHP, while Wiz independently discovered and reported both vulnerabilities, and PHP's maintainers led the fixes and release. OpenAI describes this as frontier AI finding the flaws before attackers did, opening a window for maintainers and security providers to prepare fixes and protections, and proposes a five-stage coordinated response framework for widely used software.

AI-generated editorial illustration: Patch the Planet: global response coordination as agents uncover critical vulnerabilities | OpenAI

Interpretation

OpenAI says its frontier models found critical PHP vulnerabilities before attackers did, giving maintainers and security providers time to prepare fixes and protections. Such flaws are usually exploited first and defenders begin preparing only after disclosure; OpenAI says the order was reversed here, producing what it calls a Defender's Window. Calif identified a critical flaw using GPT‑5.6 Sol as part of Patch the Planet, Wiz also independently discovered and reported both vulnerabilities, and PHP's maintainers led the fixes and release.

The response was not left to maintainers alone but divided among maintainers, researchers, cloud and security providers, and government agencies. A single maintainer team usually reviews security reports alongside routine development and cannot simultaneously adapt fixes for older releases, prepare mitigations and coordinate updates across organizations. OpenAI contributed research, technical support and coordination with partners including Wiz and Cloudflare, and engaged the White House's Super Intelligence Force, ONCD, CISA and international partners so guidance reached critical organizations.

OpenAI proposes a five-stage process that carries a finding through to a deployable fix. The stages are validating and fixing the upstream vulnerability, giving affected organizations time to act, coordinating with government partners, preparing updates across affected products, and publishing guidance and supporting deployment once a patch exists. The process weighs an organization's protective role, the time it needs to act and its ability to handle sensitive information, with maintainers deciding whom to involve ahead of disclosure.

Perspective

The report covers OpenAI's public account of the PHP vulnerability discovery and coordinated response and its proposed five-stage disclosure framework; for organizations using PHP, the stated priority is applying the fixes, and OpenAI offers a PHP version checker as a Codex skill.

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; it also says the same frontier capabilities can help attackers rapidly develop offensive tools and exploits, and that it is calibrating additional blocking classifiers to reduce misuse while acknowledging safeguards can block legitimate security work.