AI existential risk is no longer only a topic for science fiction. Researchers worked with the Existential Risk Observatory and the Future of Life Institute to map 95 scenarios in which AI poses an existential risk. They identified eight risk categories, spread over three fundamental threats.
What are existential risks in this study? They are the scenarios that could shape the future of humanity. In a unilateral takeover, one dominant AI system takes control. In a distributed takeover, several AI systems gain power independently or together. In a gradual handover, people delegate more and more decisions to AI until taking back control becomes impossible.
Other categories focus on how people use AI. Permanent totalitarianism means AI allows a small power elite to enforce authoritarian control for good. Economic inequality arises when large-scale automation concentrates economic and political power in a limited group.
The last two categories are catastrophic technology and autonomous warfare. In the first, AI speeds up technologies that could destroy humanity. In the second, AI-driven military systems cause escalations that nobody can control. Of the 95 scenarios, 56 involve an AI takeover.
The scenarios are not predictions. They show why AI safety is not only a technical issue, because human control, power distribution and international governance are essential too. The study gives policymakers, AI developers and regulators a framework to spot and limit risks early. Good AI governance starts with knowing who keeps control over the decisions you hand to AI.