OpenAI and the Frontier Model Forum Release a Framework for Rating Cyber Jailbreaks
We partnered with @fmf_org and its member firms on a new framework to help identify when bypassing an AI system's safeguards could enable serious cyber harm, and prioritize stronger protections. "...This 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. Shared approaches like this can help the AI and security communities build a clearer evidence base and prioritize improvements to safeguards." - OpenAI's Sasha Baker, Head of National Security Policy
Context
A shared way to grade jailbreaks helps labs decide which safeguard gaps to fix first.
OpenAI partnered with the Frontier Model Forum and its member firms on the framework. It aims to identify when bypassing an AI system's safeguards could enable serious cyber harm. OpenAI's Sasha Baker called it a step toward a common method for comparing jailbreak severity.