After the European Union, the Organisation for Economic Co-operation and Development (OECD) has now adopted key principles to guide artificial intelligence (AI). In total, 42 countries (36 member countries and 6 others) signed a joint text on May 22 that states that AI systems must be "robust, safe and secure throughout their life cycle”.
The United States is among the signatories, but not China. Within the European Union, not all Member States are on the list: Bulgaria, Cyprus, Croatia and Malta are missing. “The Recommendation is open to accession by non-OECD members, which would demonstrate the global relevance of the OECD's work on AI-related policies and help to give substance to its call for more international cooperation", the text states.
The document sets out five complementary values-based principles that lay the foundation for a responsible approach to supporting trustworthy AI and calls on AI stakeholders to promote and implement them. These principles are: inclusive growth, sustainable development and well-being; human-centred values and equity; transparency and accountability; robustness, safety and security; and accountability. The document also sets out five recommendations that decision-makers are invited to follow in the development of their national policies and international cooperation in order to move towards trustworthy AI: investing in AI research and development; fostering the development of a digital ecosystem for AI; shaping a policy framework supportive of AI; building human capacity and preparing for labour market transformation; and fostering international cooperation in the service of trustworthy AI.
It should be recalled that at the European level, the group of experts on artificial intelligence published seven key principles and an evaluation grid for professionals regarding the ethical dimension of AI on 9 April. On the previous day, the European Commission endorsed these seven key principles in a communication while specifying that it was giving itself a year to consider an alternative to self-regulation (see EUROPE 12231/10).
Link to the recommendation: https://bit.ly/2VMUCRW (Sophie Petitjean)