The Future of AI: Five Scenarios for a Transforming World
Summary
The Centre for the Governance of AI presents five plausible futures for advanced AI rather than a single forecast. The scenarios are built around two uncertainties: whether AI-automated research drives rapid progress toward increasingly general-purpose agents or progress plateaus into narrower applications, and whether development is concentrated in one dominant actor or distributed among competing actors. Combining those variables produces five core scenarios with ten possible endings: Plateau, Big AI, Diplomacy, Arms race, and Take-off. The study argues that a non-disruptive “business as usual” path is unlikely because even current systems can reshape work, cultural institutions, and political power. It highlights the possibility that AI agents could accelerate research by coding, running experiments, and improving training, while also warning that reward hacking, weak verification, and insufficiently scalable safeguards could outpace human oversight. The analysis says control over chips, models, data centres, cloud infrastructure, and distribution could concentrate economic and geopolitical power, particularly disadvantaging Europe. Open-weight systems could broaden access, support local and specialised models, and reduce dependence on dominant firms, but they could also increase misuse risks if safeguards are bypassed. The scenarios explore outcomes including an AI winter with practical, widely available tools; corporate consolidation and labour displacement; international safety coordination after a serious incident; military escalation between the United States and China; and self-improving systems that either enable abundance or escape meaningful human control. The authors present these as structured thought experiments grounded in current trends and uncertainties, not predictions. They recommend using scenario analysis to stress-test policy and building resilience across cybersecurity, economic institutions, information ecosystems, supply chains, and international governance.