OpenAI and FMF Release New Framework for Rating Advanced Network Threat Jailbreaks

OpenAI has partnered with the Frontier Model Forum (FMF) and its member companies to launch a new framework for assessing when bypassing AI safety measures could lead to serious cyber harm, enabling them to prioritize strengthening defenses accordingly. Sasha Baker, OpenAI's head of national security policy, stated that this framework represents a significant step toward establishing a shared methodology for evaluating and comparing the severity of AI jailbreaks, including their unlocked capabilities and availability.