Policy & society

International AI Safety Report 2026

An international scientific assessment of general-purpose AI capabilities, risks and mitigation approaches.

Publisher: International AI Safety Report
Publication: February 2026
Length: 221 pages
Language: English

£799 GBP

PDF download and private email delivery after payment. Purchases are not open yet.

Published by the named institution. Delivered as an English PDF file. Download and save your copy within 7 days of payment.

International AI Safety Report

2026 / RESEARCH

Independent research index / editorial cover

About this report

A scientific synthesis of emerging risks at the frontier of general-purpose AI. The report reviews current and possible future capabilities, malicious use, reliability failures, loss of control, labor-market effects and human autonomy, then examines technical and institutional approaches to risk management. Its broad expert input and attention to uncertainty make it useful for risk briefings and governance discussions. Read the evidence and limitations together: the report maps what is known, disputed and difficult to measure rather than offering a product-level safety certification.

The PDF is in English. This introduction and content guide are available in your selected language.

Who it is for

Policy and governance teams
Security and risk practitioners
Researchers and analysts

Inside this report

General-purpose AI and capability trajectories
Malicious use, influence and security risks
Reliability failures and loss of control
Systemic effects on work and human autonomy
Risk practices, safeguards and societal resilience

KEEP EXPLORING

International AI Safety Report 2026