OVERCOMING SKYNET: Controlling Authority, Not Intelligence, in the Age of AI

Overcoming Skynet: Controlling Authority, Not Intelligence, in the Age of AI argues that governing artificial intelligence requires more than making models behave well. We must also control what AI systems can access, authorize, and change.The book separates intelligence from authority. A system's ability to reason does not automatically justify giving it control over money, infrastructure, medical decisions, or other consequential activities. Effective governance requires clear permissions, independent enforcement, human accountability, and practical ways to interrupt or revoke authority.Across 24 chapters, the book moves from the fears shaped by science fiction to the architecture and institutions needed for responsible AI. It explores identity, least authority, enforceable policies, agent delegation, containment, oversight, regulation, and public trust. It also extends Ernesto Couso's Adapt-Optimize-Deliver methodology-originally developed for cloud adoption-into a proposed cycle for understanding an AI system's environment, improving its controls, and learning from deployment.Its central message is one of informed courage: we can pursue more capable intelligence while building stronger boundaries around its use. Progress depends on continued learning and on ensuring that responsibility remains with the people and institutions that grant machines the power to act.