International AI Safety Report

The AI Safety Report is jointly written by UK AI Security Institute and MILA – Quebec Artificial Intelligence Institute

Abstract: 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. 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 gives decision-makers a shared global picture of AI’s risks and impacts, serving as the authoritative reference for governments and organisations developing AI policies worldwide. It is already shaping debates and informing evidence-based decisions across research and policy communities.

First Update: Capabilities and Risk Implementations

This Key Update covers major breakthroughs in AI capabilities since the publication of the last Report in January 2025 and some implications for major risks. New training techniques that allow AI systems to use more computing power have helped them solve more complex problems, particularly in mathematics, coding, and other scientific disciplines. These capability improvements also have implications for multiple risks, including risks from biological weapons and cyber attacks, and pose new challenges for monitoring and controllability.

Please click on this link to read the first update.

Second Update: Technical Safeguards and Risk Management

This Key Update examines developments in technical approaches to general-purpose AI risk management, from training models to refuse harmful requests to watermarking AI-generated content. Since the publication of the 2025 International AI Safety Report, the number of companies publishing Frontier AI Safety Frameworks has more than doubled, and researchers have refined techniques for training safer models and detecting AI-generated content. However, significant gaps remain: sophisticated attackers can often bypass current defences, and the real-world effectiveness of many safeguards is uncertain.

Please click on this link to read the second update.

Image Credit: Image by Freepik