Back
International AI Safety Report 2025
webinternationalaisafetyreport.org·internationalaisafetyreport.org/publication/international...
This is the first major intergovernmental-style scientific report on AI safety, often compared to the IPCC; highly relevant for understanding the international policy landscape and current scientific consensus on AI risk.
Metadata
Importance: 85/100organizational reportreference
Summary
A landmark international scientific assessment co-authored by 96 experts from 30 countries, providing a comprehensive overview of general-purpose AI capabilities, risks, and risk management approaches. It aims to establish shared scientific understanding across nations as a foundation for global AI governance. The report covers topics including capability evaluation, misuse risks, systemic risks, and mitigation strategies.
Key Points
- •Collaborative effort by 96 experts from 30 countries, analogous to the IPCC model applied to AI safety
- •Covers the full landscape of general-purpose AI risks: misuse, misalignment, structural/systemic risks, and accident risks
- •Reviews capability assessment methodologies including benchmarks and red-teaming approaches
- •Examines potential risk management techniques spanning technical, institutional, and governance interventions
- •Intended to inform international policy coordination and provide a scientific baseline for AI safety discussions
Review
The report represents an unprecedented international collaborative effort to systematically analyze the current state and potential risks of general-purpose AI. Its key contribution is providing a nuanced, evidence-based overview of AI capabilities, potential risks across malicious use, malfunctions, and systemic impacts, and nascent risk management techniques. The report notably highlights the significant uncertainty surrounding AI development, with experts disagreeing on the pace and implications of capability advances. The methodology involves synthesizing current scientific research, incorporating perspectives from a diverse international expert panel, and providing a balanced assessment that acknowledges both potential benefits and risks. The report's strengths include its comprehensive scope, international collaboration, and transparent acknowledgment of scientific uncertainties. Key limitations include the rapid pace of AI development, which means the report's findings could quickly become outdated, and the inherent challenges in predicting complex technological trajectories.
Cited by 18 pages
| Page | Type | Quality |
|---|---|---|
| Persuasion and Social Manipulation | Capability | 63.0 |
| Heavy Scaffolding / Agentic Systems | Concept | 57.0 |
| Provable / Guaranteed Safe AI | Concept | 64.0 |
| AI Risk Critical Uncertainties Model | Crux | 71.0 |
| AI Risk Feedback Loop & Cascade Model | Analysis | 59.0 |
| AI Safety Intervention Effectiveness Matrix | Analysis | 73.0 |
| AI Risk Interaction Matrix | Analysis | 65.0 |
| AI Governance Coordination Technologies | Approach | 91.0 |
| Corrigibility | Research Area | 59.0 |
| Evals-Based Deployment Gates | Approach | 66.0 |
| AI Evaluations | Research Area | 72.0 |
| AI Safety Field Building and Community | Crux | 0.0 |
| AI Safety Intervention Portfolio | Approach | 91.0 |
| Pause Advocacy | Approach | 91.0 |
| AI Risk Public Education | Approach | 51.0 |
| AI Safety Cases | Approach | 91.0 |
| Corrigibility Failure | Risk | 62.0 |
| Optimistic Alignment Worldview | Concept | 91.0 |
Cached Content Preview
HTTP 200Fetched Apr 9, 202698 KB
International AI Safety Report 2025 | International AI Safety Report
Skip to main content
Open Table of contents
Table of contents
Contributors
CHAIR
Prof. Yoshua Bengio , Université de Montréal / Mila - Quebec AI Institute
EXPERT ADVISORY PANEL
This international panel was nominated by the governments of the 30 countries listed below, the UN, EU, and OECD.
Australia: Bronwyn Fox, the University of New South Wales
Brazil: André Carlos Ponce de Leon Ferreira de Carvalho, Institute of Mathematics and Computer Sciences, University of São Paulo
Canada: Mona Nemer, Chief Science Advisor of Canada
Chile: Raquel Pezoa Rivera, Universidad Técnica Federico Santa Maria
China: Yi Zeng, Chinese Academy of Sciences
European Union: Juha Heikkilä, European AI Office
France: Guillaume Avrin, National Coordination for Artificial Intelligence
Germany: Antonio Krüger, German Research Center for Artificial Intelligence
India: Balaraman Ravindran, Wadhwani School of Data Science and AI, Indian Institute of Technology Madras
Indonesia: Hammam Riza, Collaborative Research and Industrial Innovation in Artificial Intelligence (KORIKA)
Ireland: Ciarán Seoighe, Research Ireland
Israel: Ziv Katzir, Israel Innovation Authority
Italy: Andrea Monti, Legal Expert for the Undersecretary of State for the Digital Transformation, Italian Ministers Council's Presidency
Japan: Hiroaki Kitano, Sony Group Corporation
Kenya: Nusu Mwamanzi, Ministry of ICT & Digital Economy
Kingdom of Saudi Arabia: Fahad Albalawi, Saudi Authority for Data and Artificial Intelligence
Mexico: José Ramón López Portillo, LobsterTel
Netherlands: Haroon Sheikh, Netherlands’ Scientific Council for Government Policy
New Zealand: Gill Jolly, Ministry of Business, Innovation and Employment
Nigeria: Olubunmi Ajala, Ministry of Communications, Innovation and Digital Economy
OECD: Jerry Sheehan, Director of the Directorate for Science, Technology and Innovation
Philippines: Dominic Vincent Ligot, CirroLytix
Republic of Korea: Kyoung Mu Lee, Department of Electrical and Computer Engineering, Seoul National University
Rwanda: Crystal Rugege, Centre for the Fourth Industrial Revolution
Singapore: Denise Wong, Data Innovation and Protection Group, Infocomm Media Development Authority
Spain: Nuria Oliver, ELLIS Alicante
Switzerland: Christian Busch, Federal Department of Economic Affairs, Education and Research
Türkiye: Ahmet Halit Hatip, Turkish Ministry of Industry and Technology
Ukraine: Oleksii Molchanovskyi, Expert Committee on the Development of Artificial Intelligence in Ukraine
United Arab Emirates: Marwan Alserkal, Ministry of Cabinet Affairs, Prime Minister’s Office
United Kingdom: Chris Johnson, Chief Scientific Adviser in the Dep
... (truncated, 98 KB total)Resource ID:
b163447fdc804872 | Stable ID: sid_Qb9H7T40fn