International AI Safety Report
governmentCredibility Rating
High quality. Established institution or organization with editorial oversight and accountability.
Rating inherited from publication venue: UK Government
This is a flagship intergovernmental AI safety document; essential reading for those working on AI governance, policy, or risk assessment, representing the most authoritative international scientific consensus on frontier AI risks to date.
Metadata
Summary
The International AI Safety Report is the world's first comprehensive scientific review of general-purpose AI capabilities and risks, led by Turing Award winner Yoshua Bengio and authored by over 100 experts from 30+ countries. It provides policymakers with an authoritative global assessment of AI's risks, emerging capabilities, and risk mitigation approaches. The 2026 edition addresses AI capabilities today, emerging risks, and technical/institutional/societal risk management measures.
Key Points
- •Led by Yoshua Bengio, authored by 100+ experts, backed by 30+ countries — the largest international AI safety collaboration to date.
- •Covers what general-purpose AI can do today, what risks it poses, and how those risks can be mitigated.
- •Key update on technical safeguards notes frontier AI safety frameworks have more than doubled, but sophisticated attackers can often bypass current defenses.
- •Available in all 6 official UN languages with an Extended Summary for Policymakers and an Executive Summary.
- •Produced by the UK AI Security Institute as an annual report, with the 2026 edition published February 3, 2026.
Cited by 1 page
| Page | Type | Quality |
|---|---|---|
| Treacherous Turn | Risk | 67.0 |
Cached Content Preview
International AI Safety Report
Skip to main content
International
AI Safety Report
2026 International AI Safety Report
2026 Extended Summary for Policymakers
About the International AI Safety Report
The International AI Safety Report is the world's first comprehensive review of the latest science on the capabilities and risks of general-purpose AI systems. The work was overseen by an international Expert Advisory Panel nominated by over 30 countries and intergovernmental organisations. Written by over 100 independent experts and led by Turing Award winner Yoshua Bengio, it represents the largest international collaboration on AI safety research to date. The Report provides decision-makers in government and beyond with a shared and authorative global picture of AI's risks and impacts.
3 February 2026
—
Annual Report
International AI Safety Report 2026
The second International AI Safety Report, published in February 2026, is the next iteration of the comprehensive review of latest scientific research on the capabilities and risks of general-purpose AI systems. Led by Turing Award winner Yoshua Bengio and authored by over 100 AI experts, the report is backed by over 30 countries and international organisations. It represents the largest global collaboration on AI safety to date.
Translated versions in the other 5 official UN languages can be found under the 'More Languages' button. The 'Extended Summary for Policymakers' can be found on the main 'Publications' page.
3 February 2026
—
Summary
2026 Report: Extended Summary for Policymakers
The Extended Summary for Policymakers provides a detailed (20 page) summary of the full 2026 Report. It includes the Report’s key findings and figures, concrete examples and a list of notable developments since the 2025 edition. It is structured around three central questions: what can general-purpose AI do today, what emerging risks does it pose, and how can those risks be mitigated?
Translated versions in the other 5 official UN languages can be found under the 'More Languages' button.
3 February 2026
—
Summary
2026 Report: Executive Summary
The Executive Summary offers a concise three-page overview of the 2026 Report’s core findings on general-purpose AI capabilities, emerging risks, and risk management approaches. It covers how AI capabilities are advancing, what real-world evidence is emerging for key risks, and progress and remaining limitations in technical, institutional, and societal risk management measures.
... (truncated, 4 KB total)181a6c57dd4cbc02 | Stable ID: ODQzZGU5MT