This article examines the European Union’s attempt to regulate artificial intelligence through the AI Act against the Union’s own constitutional standard of the rule of law. It advances a functional understanding of the rule of law: not as a symbolic goal, but as a means of protecting rights by providing legality, foreseeability, control of discretion, equality and access to justice. Artificial intelligence poses a structural challenge to this grammar, as autonomy and opacity relocate decision-making from promulgated norms to model design, data selection and inference processes that are difficult to scrutinise. Using the Venice Commission’s Rule of Law Checklist as an operational benchmark, the article evaluates the AI Act element by element. It finds that legality is anchored in an internal market basis rather than in rights-oriented Treaty provisions; legal certainty is displaced into shifting Commission guidelines; prevention of abuse is compromised by reliance on delegated acts and industry co-regulation; equality is left to self-assessment through the Fundamental Rights Impact Assessment; and access to justice is nominal, with explanation rights narrow and evidence disclosure absent. The cumulative effect is constitutionalisation in form but not in substance, risking a transformation of the rule of law into rule by law. The article concludes by outlining de iure condendo reforms that could align the EU AI architecture with its constitutional commitment to the rule of law as a means of protecting rights.