GoKawiilFrontier Model Forum publishes v1.0 Jailbreak Rating Framework for AI models
The Frontier Model Forum, backed by Anthropic, Google, Microsoft and OpenAI, has released a draft scoring system for rating how severe and exploitable AI jailbreaks are. The framework scores jailbreaks on capability gain, scope and ease of deployment, combining them into a single low-to-critical rating, and is open for public comment until December 21.