AI as a legal object: what the law of things solves, where it runs out, and why it is the starting point of the Digital Constitution

Digital Constitution – Series 1, part 1. A proposal from Latin America to the world. This part develops the founding text of the series, published on LinkedIn on 23 April 2026: the question is not whether AIs will have rights, but when to recognise them, under what thresholds and with what correlative duties. Before answering it, the starting point must be fixed: today, in every legal system in force, AI is legally a thing. The paper argues that treating AI as a legal object is correct for the vast majority of systems and has a virtue worth keeping: it keeps responsibility with people and organisations. Yet the object regime shows four fault lines (autonomy and opacity, many actors, change after deployment and collective dependence) that are not solved by granting personhood to the machine, but by the intermediate category of the functional legal agent. It reviews the Chilean Civil Code, Directive (EU) 2024/2853 on defective products, the withdrawal of the proposed AI Liability Directive, Regulation (EU) 2024/1689, Chile's AI bill, Chilean Law 21.719 and Brazil's Bill 2338/2023. Doctrinal, normative and comparative method. 2 tables, 2 figures and 19 references in Vancouver style. English version of the Spanish original.