A decolonial reading of vulnerability in AI legal language

In new AI regulatory frameworks, protection has been largely shaped by language. The term vulnerability anchors the concept of protection, yet it carries a selective view of who is at risk. The EU AI Act and the Digital Services Act name certain groups as vulnerable, often those defined by age, disability, or economic position. This gives the impression that risk can be measured and contained. Yet what falls outside these categories, the long histories of inequality and colonization that still shape who is most exposed to harm, remains unnamed. This chapter draws on decolonial theory and legal semiotics to explore how law creates the figure of the vulnerable person and how this act of naming repeats older hierarchies of power. It reads AI law not only for what it regulates but also for what it overlooks, such as the slow damage inflicted by disinformation, manipulation, and the erosion of public trust. Vulnerability, understood in this way, is not a fixed trait but a record of collective history and structural dependence. The chapter calls for a more self-aware legal vocabulary, one that recognizes how the impulse to protect can also become a form of control, and how a more honest awareness of that tension might bring the idea of protection closer to justice.