On 10 September 2024, the Organisation for Economic Co-operation and Development (OECD) closes a public consultation on establishing risk thresholds for advanced Artificial Intelligence (AI) systems. The consultation focuses on the potential approaches, opportunities, and limitations in establishing risk thresholds for advanced AI systems. The consultation focuses on AI systems that may pose significant risks and seeks to gather information on thresholds that could trigger actions when these systems present unacceptable risks, whether based on computational power, system capabilities, or other factors. The consultation aims to gather information on the usefulness of current resources, the effectiveness of compute power-based thresholds, the value of alternative thresholds, strategies for setting and evaluating thresholds, and requirements for systems that exceed them.
Original source