OpenAI and Frontier Model Forum launch cyber jailbreak severity framework
OpenAI and the Frontier Model Forum announced a joint framework for identifying cyber jailbreaks that could bypass AI safeguards and enable serious harm, according to OpenAI. The framework offers a common methodology for comparing jailbreak severity based on the capabilities they unlock and how readily they can be used, and is intended to help prioritize stronger protections.
OpenAI says the shared methodology can help AI and security communities build a clearer evidence base and prioritize improvements to safeguards.
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
https://www.frontiermodelforum.org/updates/introducing-the-fmf-jailbreak-rating-framework-for-frontier-cyber-threats/