OpenAI and FMF unveil jailbreak rating framework for cyber risk
According to Frontier Model Forum, OpenAI and FMF released a framework to rate the severity and usability of AI jailbreaks to guide stronger safeguard priorities.
SourceAnalysis
OpenAI partnered with the Frontier Model Forum and its member firms to release a new framework that identifies when bypassing an AI system's safeguards could enable serious cyber harm and helps prioritize stronger protections.
The framework supports a common methodology for evaluating and comparing the severity of cyber jailbreaks, including the capabilities they unlock and how readily they can be used, according to OpenAI. Sasha Baker, Head of National Security Policy at OpenAI, stated that shared approaches like this can help the AI and security communities build a clearer evidence base and prioritize improvements to safeguards.
OpenAI
@OpenAILeading AI research organization developing transformative technologies like ChatGPT while pursuing beneficial artificial general intelligence.