Inteligência artificial e danos: autonomia, riscos e solidariedade

Abstract

Inserted in the context of the Fourth Industrial Revolution, Artificial Intelligence has been shaking the world. Its applications spread to almost all sectors of life in society, promoting significant advances in various areas, becoming the object of study of various sciences, including Law. The present dissertation, thus, arises to address only one of so many concerns: the civil liability for the damages caused by Artificial Intelligence. It seeks, therefore, to analyze this technological phenomenon and its main characteristics and techniques, such as machine learning, which enable, each day, to provide robots and intelligent systems with a growing, albeit purely technological, autonomy. It moves, then, to an investigation of two types of damage caused by Artificial Intelligence: autonomous cars and discrimination caused by biased and prejudiced algorithms. Inserted in a context of sharing economy, autonomous cars come along with the promise of a reduction in the number of accidents compared to human drivers, although they do not eliminate them and cause new specific damage. Artificial Intelligence is also increasingly employed to support automated decisions based on techniques such as data mining and profiling. Nevertheless, a number of damages can arise, since algorithms often inherit biases and prejudices from their programmers and the databases that feed them. Such unjustified discrimination may constitute autonomous harm or give rise to rights such as the right to explanation and the right to review such decisions made based solely on automated processing of personal data. As far as civil liability is concerned, in both cases the theory of fact of the product or service provided for in the Brazilian Consumer Protection Code applies in general. Nevertheless, this shall be complemented by objective and subjective imputation schemes, depending, for example, on the reciprocity of risks, in the case of autonomous cars. Civil liability for the autonomous acts of Artificial Intelligence is based, in fact, on its typology and its greater or lesser autonomy. However, it has been argued that traditional norms are not able to cope with the increased risks and damages caused by the increasingly autonomous acting of Artificial Intelligence. Therefore, the present study is dedicated to analyzing possible constructions of lege ferenda, based on the solidarity and risk-sharing of technological development, which translate into alternative and complementary solutions to Civil Liability, such as the creation of na e-personality, compulsory insurance and compensatory funds. This is based on the premise that no matter how autonomous Artificial Intelligence robots and systems are, ultimately, they will always be at the service of some human interest, which must withstand the risks generated to a greater or lesser extent, in order to realize the main purpose of the current Civil Liability system: the full reparation of the damages caused to the human person.