AI finds critical PHP flaws, triggers global patch response
OpenAI says frontier AI models uncovered two critical vulnerabilities in PHP before attackers likely could, launching a coordinated worldwide defense effort.

An AI system that scans website code to find security weaknesses so patches can be released before hackers exploit them.
- AI scans PHP code and finds hidden weaknesses
- Global team coordinates fixes before public release
- Guidance and tools are shared to protect websites
AI uncovers critical PHP vulnerabilities
OpenAI reports that Calif, an internal team using GPT‑5.6 Sol as part of its Patch the Planet initiative, identified a critical flaw, with Wiz also independently discovering and reporting both vulnerabilities 1. Wiz, a separate security firm, also independently found and reported both vulnerabilities .
A coordinated global response
PHP maintainers led the fixes and release process 1. OpenAI provided security expertise and model access to test patches and prepare guidance . Partners including Cloudflare, Wiz, government agencies — among them CISA and the White House's Super Intelligence Force — and international collaborators all contributed to the response . The coordination aimed to prepare protections before the vulnerabilities were made public .
Who came together to fix PHP vulnerabilities
- OpenAI / Califuncovered the flaws
- Wizindependently reported both
- PHP maintainersled the fixes
- Cloudflarejoined the response
- CISA & White Housegovernment partners
- International collaboratorsglobal coordination
Roles described in the story; no outside sources used
Why it matters
The vulnerabilities are among the most impactful security issues OpenAI has encountered 1. However, the company acknowledges the same AI capabilities could also help attackers rapidly build offensive tools and exploits .
What's being done next
OpenAI released a PHP version checker as a Codex skill to help operators identify vulnerable installations 1. It also proposed a five-stage framework for responding to critical vulnerabilities in widely used software during the AI era: validate and fix upstream, give organizations time to act, coordinate with government partners, prepare updates across products, and publish guidance after disclosure . The company committed $1 billion in subsidized access to its Daybreak models through a program for frontline defenders with limited security resources .
Looking ahead
These bugs are just some of several significant vulnerabilities OpenAI has found and is working to fix, and the company expects similar findings to increase sharply over the coming months 1. OpenAI is asking the security community to help improve its proposed response framework as AI capabilities continue to advance .
Corrections and updates
- Corrected on 22 October 2026. An earlier version said millions of websites could be affected; the source describes PHP as widely used and the flaws as severe but does not cite a specific number of websites affected.
Key facts
- OpenAI and Wiz independently discovered two critical PHP vulnerabilities using AI-assisted research
- Frontier AI models found the flaws before attackers, creating a Defender's Window
- OpenAI committed $1 billion in subsidized Daybreak model access for frontline defenders
- A five-stage coordination framework was proposed for future vulnerability responses
- Expectations are that similar AI-discovered vulnerabilities will increase sharply in coming months
Sources
If this story concerns you or your company, or uses something that belongs to you, ask for a correction or for it to be taken down. How requests work
Keep reading
The weekThis Week in AI New Models Flood In and Agents Get SaferHerald of AI
This Week in AI New Models Flood In and Agents Get Safer
Big open-weight releases, speed breakthroughs, and a push to make AI agents more reliable and secure.
How it worksOpenAI Releases 372 AI-Generated Math Proofs on GitHubHerald of AI
OpenAI Releases 372 AI-Generated Math Proofs on GitHub
The company is bypassing traditional journals, betting that formal verification can handle the flood of AI-produced mathematics.
OpenAI’s ChatGPT Wallet lets AI agents shop without seeing your card number
It marks the start of a new way to buy things online — your AI helper picks and pays, but you still approve every transaction and your real card stays hidden.