Artificial Intelligence, damages and liability: ethics and Law

Abstract

This article has as its research problem the following question: how ethics and law protection should deal with the damages caused by an artificially intelligent being? The aim is to verify the sufficiency of the ethical guidelines and the Brazilian legal system of civil liability in the face of the acts caused in the analysis. We oppose: fault; strict; vicarious; and no damages liabilities. Therefore, Artificial intelligence is identified as a tool, and cannot hold the duty to indemnify. Furthermore, we present critical, etymological, statistical, and legal considerations about the concept of risk inherent to strict liability. As a partial result, it is verified that the legal conception of risk is restricted to the capacity to cause damage. Based on this result, it is analyzed whether activities involving the programming of artificially intelligent entities are considered risky. Due to the uncertainty and randomness inherent in the programming, execution, and development of artificially intelligent beings (with a focus on Machine Learning and Deep Learning), the most appropriate legal protection is a fault liability due to the potential damage of artificial intelligence systems. Furthermore, we conclude that ethical guardianship can be seen as a liability exclusion mechanism to break the causal link since it is exercised when the autonomous system is developed by programming the abstention of actions that can potentially cause damage. The development of these proposals is anchored on the hermeneutic-concretive, deductive, and integrated review methods, as well as on the case study and bibliographical research techniques.