Artificial Intelligence and Redress for Damages

Abstract

Artificial intelligence has innumerable applications in society. Algorithms have a certain degree of autonomy in their functioning. Therefore, their “behaviour” evolves over time, and the relationship of cause and effect, as regards causation of damages, may be not linear as we believe. Results of AI activity may be unpredictable despite no flaw in design or implementation. Many proposals were made in order to adapt civil liability rules in this. The author drafts here a proposal grounded on the ideas that (1) overall benefits of artificial intelligence evolution outweigh costs deriving therefrom, so that it should be encouraged or, at least, not hindered; (2) “traditional” civil liability rules (either based on fault or strict liability) may provide a negative incentive toward such evolution, insofar as they may impose the obligation to pay redress onto producers and programmers of AI devices despite no flaw in design or implementation. He proposes, in this respect, no-fault redress schemes as an interesting and worthy regulatory strategy to this end.