<p>This paper addresses the question of whether legal judgement can be outsourced to artificial intelligence (AI), and specifically to AI systems based on Large Language Models (LLMs). The issue has attracted international attention, both at the level of theoretical research and of practical application. Moreover, the development of LLM tools specifically designed for lawyers and judges (Legal LLMs) has sparked further discussions. Several scholars have expressed scepticism and highlighted limitations and concerns, while others have pointed out how operationalisation in the legal field seems to be inevitable and even desirable, once appropriate regulation is devised. This paper, however, approaches the issue from a different angle. It turns to the philosophy of Immanuel Kant in order to reconsider the very notion of judgement and, in particular, to examine what kind of judgement is at play in legal proceedings. Drawing on Kant’s distinction between <i>determinative</i> and <i>reflective</i> judgement, the analysis focuses on the specifically Kantian account of how judgement proceeds syllogistically. Particularly, it analyses the determinative mode of judgement, where the universal is already given and the particular subsumed under it, and the reflective mode, where, by reflecting on the particular, the universal is still to be sought. The paper compares these Kantian features of judgement with the current capabilities of LLMs, ultimately asking whether legal judgement, understood in both its determinative and reflective dimensions, can indeed be outsourced to them.</p>