OpenAI Patch the Planet: Global Response Coordination for Critical PHP Vulnerabilities

AI-Driven Discovery and Global Coordination of PHP Vulnerabilities

OpenAI, in collaboration with researchers and security providers, led a coalition to address critical vulnerabilities in PHP, an open-source programming language. A critical flaw was identified using GPT-5.6 Sol as part of the Patch the Planet initiative, and the vulnerabilities were independently discovered and reported by Wiz.

Given PHP’s widespread use, this incident is considered one of the most impactful security issues in recent memory. The discovery highlights AI's capability to uncover security flaws with internet-scale impact, while also underscoring the risk that the same capabilities could be used by attackers to rapidly develop exploits.

Crucially, frontier AI found these vulnerabilities before attackers did, creating a Defender’s Window that allowed maintainers and security providers time to prepare fixes and protections ahead of public disclosure.

For organizations using PHP, the immediate priority is applying the available fixes. OpenAI has provided a PHP version checker as a Codex skill to help operators identify vulnerable versions and determine which updates are applicable.

The Five-Stage Coordinated Response Framework

To manage the complexity of patching widely used software, OpenAI proposes a five-stage framework for responding to critical vulnerabilities in the AI era. This repeatable process helps teams act quickly as new vulnerabilities are discovered.

1. Validate and Fix the Upstream Vulnerability

The maintainer, researchers, and essential technical partners confirm the finding, assess its impact, and develop and test a fix. This stage includes adversarial testing using frontier models to identify potential workarounds and related variants of the vulnerability.

2. Give Affected Organizations Time to Act

A framework is proposed for determining which organizations require advance information to prepare protections. This involves outreach, information-sharing agreements, and coordination between teams. Actions can include a distributor adapting a fix for an older release, a hosting provider testing a mitigation, or a security provider preparing detection measures like Web Application Firewall (WAF) rules.

3. Coordinate with Government Partners to Protect Critical Systems

Government agencies are essential for coordinating a wider response and ensuring security guidance reaches organizations managing critical infrastructure. In the PHP case, OpenAI and its partners engaged multiple parts of the U.S. government, including the White House’s Super Intelligence Force, the Office of the National Cyber Director (ONCD), and the Cybersecurity and Infrastructure Security Agency (CISA), alongside international partners.

4. Prepare Updates Across Affected Products

Once the fix, guidance, and disclosure plans are stable, maintainers involve additional distributors, hosting and service providers, response teams, and operators to prepare necessary updates and address coverage gaps.

5. Publish Guidance and Support Deployment

Public guidance allows users to assess their exposure, find authoritative updates or interim mitigations, and verify that a fix is running. This phase continues as operators deploy updates and report any required assistance.

Scaling Defender Capacity and Resources

Global coordination relies on supporting the small teams that maintain open-source software. OpenAI is investing in models, tools, and partnerships to help defenders leverage advanced AI capabilities for vulnerability discovery, system monitoring, and threat response.

OpenAI’s investments include:

Codex for OSS: Providing subscriptions for all maintainers.
Direct Sponsorship: Sponsoring projects like the Rust Foundation.
Hands-on Support: Open-source projects can apply to Patch the Planet for direct engineering assistance.

Furthermore, OpenAI has committed $1 billion in subsidized Daybreak access through Daybreak for Frontline Defenders, focusing on organizations with limited security resources protecting essential services.

OpenAI is also refining blocking classifiers for Daybreak to balance the need for broad access to capable models for defenders against the necessity of limiting harmful misuse.

Building More Secure Software Foundations

Beyond fixing individual bugs, the long-term goal is to help engineers build more secure software from the initial design phase. AI can assist in several key areas to prevent vulnerabilities or limit their impact:

Secure Code from the Start: Coding models can help developers choose safer defaults, use security libraries correctly, and identify risky patterns during development.
Memory-Safe Software: AI can assist with migrating critical components to memory-safe languages, supported by tests and expert review to ensure behavioral compatibility.
Formal Verification: Mathematical proofs can be applied to widely used components to demonstrate that they satisfy specified security properties, ruling out entire classes of vulnerabilities.
Stronger Isolation: Running software that processes untrusted files in tightly restricted sandboxes limits the damage if a vulnerability is exploited. AI can help test and strengthen these sandboxing technologies.

Seeking Input to Improve the Approach

OpenAI is soliciting feedback from maintainers, researchers, response teams, providers, and operators to refine the proposed vulnerability response framework. Specific areas of interest include:

Disclosure Windows: Determining the necessary time for maintainers and operators to prepare and deploy reliable fixes, and what evidence should trigger earlier disclosure.
Emergency Patching: Defining safeguards for when governments or operators must use a vulnerability to patch systems after losing administrative access.
Early Notice: Establishing who needs advance warning for protective action, and how to balance coverage against the risk of a leak.
Secure-by-Default Packages: Finding ways to help identify and fix vulnerabilities in downstream packages while upstream fixes are pending, or backporting fixes to older, unsupported versions.