Human Responsibility and Control over Artificial Decision-making in Italy: Challenges and Problems

ABSTRACT The essay examines the principle of human responsibility and control over automated and AI-assisted decision-making in Italian public administration, building on the framework of ‘algorithmic legality’ developed by administrative case law and now partly codified in Article 30 of the Public Contracts Code (Legislative Decree No. 36/2023) and in the new national AI Law (Law No. 132/2025). After outlining the constitutional and European (GDPR, AI Act) roots of the non-exclusivity/human-in-the-loop principle, the essay questions whether human oversight, as currently framed, can remain effective as AI systems grow more complex and less interpretable. It identifies several structural risks – merely formal or ‘rubber-stamp’ supervision, automation bias, fragmented responsibility among multiple actors, and insufficient administrative skills – that may turn human-in-the-loop into a ‘phantom guarantee’. Drawing a comparison with the deregulatory turn taken by the United States, the essay argues that Italian and European law nonetheless preserve an irreducible space for the human official, rooted in the personalist and impartiality principles of the Constitution. It concludes by proposing the model of the ‘artificial co-pilot’ as a viable path to reconcile the efficiency gains of AI with the protection of individual rights, provided it is accompanied by adequate investment in administrative capacity and governance.

KEYWORDS Algorithmic legality - Human-in-the-loop - Artificial intelligence - Public administration - Automation bias - Reserve of humanity

TABLE OF CONTENTS 1. Introduction: Technological Change and the Maintenance of Legal Principles. – 2. Human Responsibility and Oversight in the New Artificial Legality. – 3. Human-in-the-Loop and the Risk of Official Capture. – 4. The Irreducible Role of the Individual and the Artificial Co-pilot.