🤖 AI Summary
The global AI safety community lacks a unified evaluation framework and empirically grounded benchmark for assessing advanced AI systems. Method: Commissioned jointly by governments from 30 countries, the United Nations, the OECD, and the European Union, this project engaged over 100 interdisciplinary AI experts to independently develop the first multilateral, internationally authoritative AI safety assessment framework. It integrates the Delphi method, cross-institutional evidence synthesis, and multi-stakeholder consensus mechanisms to systematically integrate technical capabilities of state-of-the-art AI systems, emerging risks—including loss of control and malicious misuse—and governance gaps. Contribution/Results: The project delivers the first comprehensive international AI safety benchmark report, structured along three dimensions—capabilities, risks, and governance. It establishes a globally endorsed risk taxonomy and quantifiable safety capability metrics, providing an evidence-based foundation and actionable reference for national and international AI governance policy development.
📝 Abstract
The first International AI Safety Report comprehensively synthesizes the current evidence on the capabilities, risks, and safety of advanced AI systems. The report was mandated by the nations attending the AI Safety Summit in Bletchley, UK. Thirty nations, the UN, the OECD, and the EU each nominated a representative to the report's Expert Advisory Panel. A total of 100 AI experts contributed, representing diverse perspectives and disciplines. Led by the report's Chair, these independent experts collectively had full discretion over the report's content.