PARIS - With the release of its Recommendation on the Ethics of Artificial Intelligence, the United Nations Educational, Scientific and Cultural Organization (UNESCO) illustrated the current tendency to establish global norms framing the development and use of artificial intelligence (AI). Even if the intentions of such a quest for a set of universal rules are good, the realization is disputable. Indeed, establishing a normative framework to avoid potential harm and make sure that AI will benefit humanity sounds legitimate, and like a goal that is worth pursuing. Yet behind this endeavor are hidden problematic preconceptions regarding the relevance of a universalist approach to the subject. In other words, even if one feel that this undertaking is laudable, one cannot always see the potential drift this could lead to.

The universalist a priori 

Current discussions on ethics applied to AI (EA2AI) - also misleadingly called ‘AI ethics’ - are pervaded with strong a priori regarding the existence of universal values on which a universal code of EA2IA could, or should, be built. Mistakenly grounded on the assumption that there are values which are truly universally shared, this is often illustrated by putting forward the Universal Declaration of Human Rights, which assumes that the very existence of such a document proves that human rights are universally cherished. Digging a bit deeper into the subject, however, quickly reveals that things are far from being as simple as they seem. Indeed, human rights themselves are not valued by every single human on earth, nor even by all human communities. A quick look at the history of the Declaration and at its implementation casts doubts on its universality, a