Governing Through Standards: Artificial Intelligence and Values

EasyChair Preprint 15179, version 2

Abstract

The upcoming European Union’s regulation on Artificial Intelligence (AI), known as the AI Act, has opened the door for the European Commission to request the development of supporting AI harmonised standards by European Standardisation Organisations (ESOs). The standardisation request will identify the areas in which ESOs are to develop standards based on the essential requirements in the AI Act. The current draft standardisation request establishes that deliverables are to take into account the policy objectives of the commission, such as ensuring that AI systems are in respect of Union values. For ESOs, this task is complicated by the diversified world-wide network of standards-developing organisations and working groups in AI. We examine the state of the art in AI standardisation, analyse how standards embed values and identify an approach that accommodates different sets of values. While currently, there is no harmonised approach to embed value consideration in AI standardisation, there is potential for an approach geared toward flexibility with space for different configurations. In the EU, the value of freedom as movement builds the basis and the need for flexible standards that enhance interoperability between regulatory contexts with different sets of values. In global terms, there is a need for a minimum threshold of agreed-upon values within AI standards that allow different configurations based on specific regulatory contexts.

Keyphrases: AI Act, AI-based systems, Artificial Intelligence (AI), Artificial Intelligence AI, European Union, European standardisation, European standards, European values, ai harmonised standards, ethical values, european telecommunications standards institute etsi, freedom as movement, future ai standards, harmonised rules on artificial intelligence, harmonised standards, human dignity freedom democracy equality, international standards, iso iec, law and respect for human rights, presumption of conformity, program for ai ethics and governance standards, regulatory context values, regulatory contexts, rule of law, standardisation, standardisation process, treaty on european union, union values, values, values of respect for human dignity freedom

@booklet{EasyChair:15179,
  author    = {Ana Paula Gonzalez Torres and Timo Ali-Vehmas},
  title     = {Governing Through Standards: Artificial Intelligence and Values},
  howpublished = {EasyChair Preprint 15179},
  year      = {EasyChair, 2024}}