In February 2020, the European Commission published a White Paper on Artificial Intelligence (AI)(1), presenting policy options to ensure trustworthy and secure development of AI, in full respect of EU values and the rights of EU citizens. It outlines ideas for a future initiative for AI in Europe and also relates to issues such as transparency and explainability.

To address the risks related to some AI applications, rather than regulate the specific algorithms being used, the White Paper proposes a risk-based approach that is meant to be technology-agnostic and future-proof. AI applications considered high-risk could be subject to a set of requirements and a prior conformity assessment.

The possible requirements could include provision of clear information of an AI system’s capabilities and limitations, and record-keeping obligations regarding the data set used to train and test the AI system. This requirement could also cover documentation on the methodologies, processes and techniques used to build, test and validate the AI system, including where relevant in respect of safety and avoiding bias that could lead to prohibited discrimination.

Other possible requirements in the White Paper specifically refer to robustness and accuracy, such as ensuring reproducibility which is essential to assess a system’s outcomes, and human oversight, adapted to the intended use and possible effects of the system, including human review of automated decisions.

The future initiative on AI is being developed following the public consultation conducted in the spring of 2020. It is expected to be adopted in the first quarter of 2021.