OpenAI and the Financial Markets Forum launch a new framework to assess risks of AI exploitations

OpenAI collaborated with the Financial Markets Forum and its member organizations to develop a new framework that helps identify when the circumvention of an AI system’s defenses could lead to serious cyber harms, and to prioritize stronger protective measures. The methodology aims to provide comparable assessments of attack severity, escaped capabilities, and exploitability, thereby creating a better evidence base and development priorities.