Causal imputation in damages derived from artificial intelligence

Abstract

This article analyzes causal attribution in damages arising from artificial intelligence (AI), focusing on the legal challenges that arise due to algorithmic autonomy and the complexity of determining material authorship. The research explores different scenarios of harm caused by AI, from those resulting from its regular use to those originating from errors in data processing or unpredictable autonomous decisions. Responsible parties, such as designers, suppliers, and operators, are discussed, highlighting the transparency, oversight, and control obligations established in regulations such as the European AI Regulation. The study evaluates the applicability of the adequate causation theory, concluding that, although useful in cases of foreseeable harm, it is insufficient in the face of erratic algorithmic behavior. It proposes complementary criteria, such as reasonable probability and objective attribution, based on permissible and prohibited risk, to address causal attribution in contexts of high uncertainty. Finally, the need for a regulatory framework that balances technological benefits with the protection of fundamental rights, ensuring civil liability in an evolving digital environment, is emphasized.