Particle.news

OpenAI Launches Patch the Planet to Help Fix Open-Source Security Flaws

It uses GPT-5.5-Cyber plus Codex Security under trusted access with Trail of Bits engineers to find, validate, and ship patches.

Overview

  • OpenAI and security firm Trail of Bits announced Patch the Planet on Monday, June 22, 2026, a program to provide free security consulting and tooling to open-source maintainers so they can find, triage, and patch vulnerabilities faster.
  • In its first week the effort engaged more than 30 projects and, the companies say, identified hundreds of bugs and produced dozens of patches while leaving maintainers with reusable tests and workflows and six months of ChatGPT Pro and Codex Security access.
  • OpenAI is offering a limited Trusted Access model of GPT-5.5-Cyber for verified defenders and updated Codex Security tooling that can scan code, trace attack paths, generate candidate patches, and run validation checks inside developer workflows.
  • Trail of Bits engineers do human validation, build and test fixes, and train project-specific automation so volunteers are not overwhelmed by low-quality AI reports and can focus on reviewing and landing vetted patches.
  • The program arrives as intelligence partners warn that powerful AI models will rapidly change cyber offense and defense, a reality that has pushed firms to restrict advanced models and could speed defensive patching while raising questions about access controls and oversight.