OpenAI launches initiative to find and patch open-source bugs with AI

OpenAI launched a new initiative to use its AI models — including a freshly released GPT-5.5-Cyber variant — to help the open-source community find and patch vulnerabilities. The program pairs OpenAI models with maintainers and platforms like HackerOne and Trail of Bits under the

The initiative is timed to a wave of AI-driven security tooling: GPT-5.5-Cyber scored 85.6% on the CyberGym benchmark (the highest single-model result to date), and Anthropic's Claude Mythos 5 is deployed under "Project Glasswing" for cyberdefenders. For developers, this means the same AI models that write their code are increasingly auditing it — and proposing patches upstream.