Risks of generative artificial intelligence systems and the European Artificial Intelligence Regulation

Abstract

Generative artificial intelligence allows the automatic creation
of “original” content in different formats (text, audio, video and
images). Content that is getting closer and closer to that created
by humans. Generative artificial intelligence offers society as
many benefits as possible risks. Some will be analysed such as
confusion, disinformation, copyright risks, security, privacy, cybercrime,
the right to honour, privacy and self-image that can be
injured by deep fakes, manipulation, risks in the labour market,
psychological risks, bias... The new Regulation of the European
Parliament and of the Council laying down harmonized rules on
artificial intelligence (Artificial Intelligence Act) welcomes generative
artificial intelligence within general purpose intelligence,
considering the presence of possible systemic risks and the enforceability
of cybersecurity. The connection of generative artificial
intelligence with prohibited and high-risk artificial intelligence
will also be analysed. The Regulation shows special concern
for “ultra-phishing”, deep fakes (where it is difficult to distinguish
between the real and the unreal), and the necessary knowledge
of their artificial origin when natural persons interact with them,
with the corresponding obligation of transparency. All this, without
forgetting the need for legal ethics and literacy.