From the “Levissima Offensa” of the pledged slave to the deterioration of AI assets given as security: between Roman Law and european regulation

This article explores the intersection of Roman law and contemporary artificial intelligence regulation through a comparative analysis of D. 20.1.27, a fragment from the classical jurist Ulpius Marcellus. The fragment concerns a debtor who pledged a slave without delivering him to the creditor and subsequently placed him in chains on account of minor misconduct, thereby diminishing the value of the secured asset. Marcellus devised a special actio utilis to protect the pledgee against such prejudice, grounding liability in the objective economic harm suffered by the creditor. Ulpian's note introduced a subjective criterion — distinguishing whether the debtor acted to harm the creditor or in legitimate response to the slave's conduct. The article argues that this Roman jurisprudential debate offers a valuable heuristic framework for addressing the modern problem of AI asset deterioration in secured transactions. Drawing on the European AI Act, it examines how obligations of technical documentation, performance monitoring and human oversight create objective parameters analogous to those developed by Marcellus, while Ulpian's subjective distinction retains residual utility. The comparison ultimately demonstrates the enduring vitality of Roman scientia iuris as a paradigm for legal systems confronting novel forms of asset and liability.