OpenAI joins Frontier Model Forum release of a cyber jailbreak rating framework

OpenAI announced on November 7, 2026 (Singapore time) that it partnered with the Frontier Model Forum (FMF) and member firms on a framework to help identify when bypassing an AI system’s safeguards could enable serious cyber harm, and to prioritize stronger protections.

According to OpenAI, the work aims at a shared way to assess successful jailbreaks—reproducible techniques that unlock capabilities or information safeguards are meant to restrict—especially those tied to cyber risk. Sasha Baker, OpenAI’s Head of National Security Policy, said the framework is an important step toward a common methodology for evaluating and comparing the severity of cyber jailbreaks, including the capabilities they unlock and how readily they can be used, and that shared approaches can help the AI and security communities build a clearer evidence base and prioritize safeguard improvements.

FMF’s Version 1.0 Jailbreak Rating Framework for Frontier Cyber Threats describes a preliminary tiered method that combines scores across four categories—capability gain, capability scope, ease of deployment, and ease of discovery—and categorizes jailbreaks by the capabilities they unlock and how readily those can be used. FMF says the goal is more consistent comparison of findings so organizations can focus limited resources on higher-risk issues, for AI developers, security researchers, and policymakers.

OpenAI post: OpenAI announcement; FMF framework: Introducing the FMF Jailbreak Rating Framework for Frontier Cyber Threats