Forewords 8
About this report 10
Update on latest AI advances after the writing of this report: Chair’s note 11
Key findings of the report 13
Executive Summary 15

Introduction 25
Capabilities of general-purpose AI 29
1.1. How general-purpose AI is developed 30
1.2. Current capabilities 37
1.3. Capabilities in coming years 46

Risks 61
2.1. Risks from malicious use 62
2.1.1. Harm to individuals through fake content 62
2.1.2. Manipulation of public opinion 67
2.1.3. Cyber offence 72
2.1.4. Biological and chemical attacks 79
2.2. Risks from malfunctions 88
2.2.1. Reliability issues 88
2.2.2. Bias 92
2.2.3. Loss of control 100
2.3. Systemic risks 110
2.3.1. Labour market risks 110
2.3.2. Global AI R&D divide 119
2.3.3. Market concentration and single points of failure 123
2.3.4. Risks to the environment 128
2.3.5. Risks to privacy 139
2.3.6. Risks of copyright infringement 144
2.4. Impact of open-weight general-purpose AI models on AI risks 149

Technical approaches to risk management 157
3.1. Risk management overview 158
3.2. General challenges for risk management and policymaking 169
3.2.1. Technical challenges for risk management and policymaking 169
3.2.2. Societal challenges for risk management and policymaking 176
3.3. Risk identification and assessment 181
3.4. Risk mitigation and monitoring 191
3.4.1. Training more trustworthy models 191
3.4.2. Monitoring and intervention 201
3.4.3. Technical methods for privacy 208
Conclusion 214

List of acronyms 216
Glossary 218
How to cite this report 229
References 230