3. Security and AI

If sovereignty determines who owns the digital ground, security determines whether it holds. This chapter extends the preceding argument into the domain of AI-era threats, where the old mantra of "trust but verify" must become "verify before you trust." It examines how automation creates a dangerous gap: responsibility stays human while authority drifts to systems that cannot explain their own decisions. Drawing on cases including jurisdictional exposure of cloud infrastructure, prompt exploitation as a form of social engineering against social machines, and cognitive warfare through poisoned training data, the chapter argues that AI security is no longer a technical discipline alone, it is governance, transparency, and the institutional habit of asking difficult questions. The chapter confronts the asymmetry at the heart of modern threats: AI agents that probe continuously, learn patiently, and apply pressure that looks not like an attack but like a perfectly timed message. Against this, it positions weight. Encryption you control, logging you can read, models you can challenge, decisions that stay human, as the necessary counterforce to a culture of convenient, unexamined lightness. Security, the chapter concludes, is not isolation; it is the discipline of keeping choice in your own hands.