In October 2023, TFS contributed to the United States and European Union’s public consultations on the G7 Hiroshima Artificial Intelligence Process Guiding Principles and Code of Conduct, which is intended to inform the actions of organizations developing the most advanced AI systems, including the most advanced foundation models and generative AI systems.

The October 30th, 2023 version of the G7 Hiroshima Artificial Intelligence Process Code of Conduct incorporated changes reflecting TFS’s recommendations to include terminology related to systemic risks, risk prevention, and interoperability across international technical standards and frameworks. Our input’s impact assessment is available upon request by contacting info@thefuturesociety.org.

Our recommendations underscored the need for a comprehensive approach to managing the risks presented by the development and deployment of AI systems. This includes emphasizing the specification and mitigation of systemic risks and major accidents, implementing stringent know-your-customer processes, and prioritizing pre-registration and notification protocols for advanced AI systems. The recommendations stress the significance of proactive prevention and monitoring of risks, especially with regard to open-source AI models.

We further advocated for the establishment of international technical standards and protocols to ensure consistent measurements of AI system capabilities and trustworthiness. To enhance accountability and transparency, the proposed modifications emphasize the need for clear roles and responsibilities for quality management, external audits, and compliance with the established principles. We proposed regular reporting on compliance, along with the facilitation of external reviews by authorized entities.

The G7 Leaders’ Statement on the Hiroshima AI Process can be accessed here, and the draft Code of Conduct can be found here (U.S. Department of Commerce update forthcoming).