October 21, 2026
Over the past two months, we have led a coalition of maintainers, researchers, and security vendors to address critical vulnerabilities in PHP—an open-source programming language that powers websites and web applications. As part of our Patch the Planet initiative, Calif uncovered a critical flaw using GPT-5.6 Sol, while Wiz independently discovered and reported both vulnerabilities. PHP maintainers led the remediation and release efforts, while our team, alongside partners including Wiz and Cloudflare, provided research, technical support, and coordination to ensure defenses were ready ahead of disclosure.
For organizations running PHP, applying these patches is an immediate priority. We are sharing a PHP version-checking tool as a Codex skill to help operators identify potentially vulnerable PHP versions and determine the appropriate updates in line with guidance from PHP maintainers and their vendors.
Given the severity of these flaws and the widespread adoption of PHP, we consider this one of the most impactful security issues in recent years. It underscores how AI can uncover internet-scale security defects—even as the same capabilities could allow adversaries to rapidly develop attack tooling and exploits.
To our knowledge, frontier AI surfaced these vulnerabilities ahead of attackers, securing a Defender’s Window for maintainers and security vendors to prepare fixes and defenses before public disclosure.
To help more defenders harness these capabilities, we have committed $1 billion in subsidized Daybreak access through Daybreak for Frontline Defenders, focusing on organizations with limited security resources that are tasked with protecting critical services.
Significant as they are, these vulnerabilities are just a subset of the critical issues we have discovered and are working to remediate, and we anticipate similar discoveries will increase dramatically in the coming months. Based on our experience to date, we are sharing a proposed framework for addressing critical vulnerabilities in widely used software in the age of AI, and we invite the security community to help refine it. Bringing together our models, security expertise, and software maintainers, this framework aims to turn discoveries into protective capabilities and build a response mechanism that scales alongside AI advancements.
Coordinating a Global Response
Deploying patches can require backporting fixes to older releases, preparing vendor mitigations, and updating systems across thousands of organizations. As more vulnerabilities are identified, a repeatable process helps these teams move swiftly.
Through Patch the Planet, we contributed security expertise and model access to support the PHP response, collaborating with maintainers and research partners to test patches and develop guidance. We also helped coordinate with infrastructure providers to protect affected systems and engaged with government cybersecurity agencies to ensure guidance reached critical organizations. This bolstered the maintainer-led response and helped prepare other defenders.
For some discoveries, researchers can work directly with software maintainers to investigate issues and develop patches. Other vulnerabilities demand a broader response because they affect numerous downstream products, or because distributors and service providers need time—a window that may be razor-thin if the gap between frontier models and widely available tools is narrow—to prepare updates and defenses. In these cases, the proposed framework helps maintainers identify who to mobilize and coordinate the efforts. This response process unfolds across five stages:
-
Verify and fix upstream vulnerabilities. Maintainers, researchers, and key technical partners validate the flaw, assess its impact, and develop and test patches. This includes adversarial testing with frontier models to look for patch bypasses and related vulnerability variants.
-
Provide lead time for affected organizations. We propose a framework to evaluate which organizations need advance notice to prepare defenses, with maintainers retaining final discretion over who participates. We assist with outreach, information-sharing agreements, and cross-team coordination. Distributors can backport patches to older versions, hosting providers can test mitigations, and security vendors can prepare detections and protections, including Web Application Firewall (WAF) rules.
-
Coordinate with government partners to safeguard critical systems. Government agencies play an indispensable role in orchestrating broader responses and ensuring security guidance reaches organizations responsible for critical infrastructure and essential services. In the case of PHP, we collaborated with partners to engage multiple parts of the U.S. government—including the White House Super Intelligence Force, the Office of the National Cyber Director (ONCD), and the Cybersecurity and Infrastructure Security Agency (CISA)—as well as international partners to support this effort.
-
Prepare updates for affected downstream products. As fixes, guidance, and disclosure timelines stabilize, maintainers can bring in additional distributors, hosting and service providers, incident response teams, and operators to prepare updates and address coverage gaps.
-
Publish guidance and support deployment following patch release. Public guidance helps organizations assess their exposure, identify authoritative updates or temporary mitigations, and verify that patches are working properly. Response efforts continue as operators roll out updates and seek assistance as needed.
The proposed framework weighs an organization’s defensive responsibilities (including protecting its own systems and its customers), the time required to take action, and its ability to handle sensitive information. The same criteria will apply across peer organizations.