International AI Safety Report
In plain wordsA large scientific review, led by the computer scientist Yoshua Bengio and backed by more than 30 countries, summing up what is known about what advanced AI can do and how it could go wrong.
A recurring scientific assessment of the capabilities and risks of general-purpose AI and of the methods for managing them, written by more than 100 independent experts under the chairmanship of Yoshua BengioYoshua Bengio, prop. n.: A Canadian computer scientist who helped found deep learning, including ideas behind today's language models, and who has become a leading voice on the risks of advanced AI. and advised by a panel of nominees from more than 30 countries and international organizations, including the EU, the OECD and the UN. Loosely modeled on the Intergovernmental Panel on Climate Change, it synthesizes evidence rather than recommending policy, and groups risks into malicious use (such as cyberattacks, biological misuse and deepfakesdeepfake, n.: A fake but realistic video, photo or voice recording made with AI that puts words in a real person's mouth or shows them in a scene that never happened.), malfunctions (including unreliability and loss of controlloss of control, n.: A situation in which people can no longer oversee, correct or switch off an AI system, and have no reliable way of getting that power back.) and systemic risks (such as labor-market disruption and concentration of power). Its secretariat sits in the UK AI Security Institute. Editions comprise an interim report (May 2024), the first full report (January 2025), shorter key updates (late 2025) and the second full report (3 February 2026).
The International AI Safety Report notes that some models behave differently when they detect that they are being tested.
Origin
Commissioned by the UK government after the Bletchley DeclarationBletchley Declaration, prop. n.: A short statement signed by 28 countries and the European Union at the first global AI safety summit, held in England in 2023, agreeing that the most advanced AI could cause serious harm. of November 2023 called for an internationally inclusive network of scientific research on frontier AI safety. The interim edition, titled International Scientific Report on the Safety of Advanced AI, was presented at the Seoul summit in 2024; later editions informed the Paris (2025) and New Delhi (2026) summits.
Usage note
Not a UN body. The Independent International Scientific Panel on AIIndependent International Scientific Panel on AI, prop. n.: A group of 40 independent experts set up by the United Nations in 2025 to give every country a shared, science-based picture of what AI can do, its risks and its benefits., created by the UN General Assembly in 2025, has a broader remit covering opportunities as well as risks; Bengio chairs the report and co-chairs the panel.
Related
- Yoshua BengioYoshua Bengio, prop. n.: A Canadian computer scientist who helped found deep learning, including ideas behind today's language models, and who has become a leading voice on the risks of advanced AI.
- Bletchley DeclarationBletchley Declaration, prop. n.: A short statement signed by 28 countries and the European Union at the first global AI safety summit, held in England in 2023, agreeing that the most advanced AI could cause serious harm.
- AI Safety SummitAI Safety Summit, prop. n.: A series of international meetings of governments, AI companies and researchers, begun in England in 2023, to agree on how to handle the risks and benefits of the most advanced AI.
- Independent International Scientific Panel on AIIndependent International Scientific Panel on AI, prop. n.: A group of 40 independent experts set up by the United Nations in 2025 to give every country a shared, science-based picture of what AI can do, its risks and its benefits.
- AI Safety InstituteAI Safety Institute, n.: A government agency that tests the most advanced AI models for dangerous abilities and studies how to make them safe. Britain and the United States set up the first ones in 2023.
- frontier AIfrontier AI, n.: The most capable general-purpose AI systems of the moment — those at or beyond the leading edge — which governments single out for extra safety attention.
Sources
- Bengio, Y. et al. (2026). International AI Safety Report 2026. internationalaisafetyreport.org/publication/international-ai-safety-report-2026
- Bengio, Y. et al. (2025). International AI Safety Report. UK Department for Science, Innovation and Technology, January 2025.
Mentioned in
AI Safety InstituteAI Safety Institute, n.: A government agency that tests the most advanced AI models for dangerous abilities and studies how to make them safe. Britain and the United States set up the first ones in 2023., AI Safety SummitAI Safety Summit, prop. n.: A series of international meetings of governments, AI companies and researchers, begun in England in 2023, to agree on how to handle the risks and benefits of the most advanced AI., loss of controlloss of control, n.: A situation in which people can no longer oversee, correct or switch off an AI system, and have no reliable way of getting that power back., Bletchley DeclarationBletchley Declaration, prop. n.: A short statement signed by 28 countries and the European Union at the first global AI safety summit, held in England in 2023, agreeing that the most advanced AI could cause serious harm., general-purpose AIgeneral-purpose AI, n.: AI that can do a wide range of different tasks, like writing, coding, translating and analyzing images, rather than the one job it was built for. Today's chatbots are the main example., Independent International Scientific Panel on AIIndependent International Scientific Panel on AI, prop. n.: A group of 40 independent experts set up by the United Nations in 2025 to give every country a shared, science-based picture of what AI can do, its risks and its benefits., Yoshua BengioYoshua Bengio, prop. n.: A Canadian computer scientist who helped found deep learning, including ideas behind today's language models, and who has become a leading voice on the risks of advanced AI., High-level Advisory Body on Artificial IntelligenceHigh-level Advisory Body on Artificial Intelligence, prop. n.: A group of experts set up by the UN Secretary-General in 2023 to advise on governing AI worldwide; its 2024 report led to the UN's AI science panel and global dialogue.