European legal framework to guarantee quality and compliance with ethical principles

On 19 February, the European Commission published a White Paper on ‘Artificial Intelligence - a European Concept for Excellence and Trust’, COM(2020) 65 final.  The White Paper extends the industrial and regional policy approach already taken to develop and promote the widespread use of AI, including the use of big data, and proposes a coordinated European approach. The White Paper identifies various policy options, including those in the fields of healthcare and public administrations. A programme for introducing AI will be developed for each sector, which will promote the adoption of AI systems and help to transform public procurement processes.

The Commission also advocates ‘a clear European regulatory framework’ that protects fundamental rights and citizens’ rights, including data protection, privacy, non-discrimination and consumer protection. In contrast, a country-specific approach would lead to fragmentation of the internal market.

The Commission provides a detailed description of the various potential risks associated with AI, including the ‘black box effect’, that is, the difficulty of tracing back actions and decisions involving the use of AI. Certain features of AI can make it difficult to apply and enforce existing legislation, such as non-discrimination. Therefore, it may be necessary to amend this legislation.  

The Commission wants to adopt a risked-based approach to adapting relevant legislation. Any AI applications that are not considered to be high-risk would continue to be fully subject to existing EU rules. The first step in determining whether an application is high-risk to identify if it is used in a sector associated with significant risks, such as healthcare, social security and employment services. The next step would then be to identify if the specific use of the AI application in this sector poses a significant risk, such as the risk of injury or a violation of rights.

Possible approaches for future regulations include improving the quality of training data (e.g. to avoid unwanted discrimination), guaranteeing the traceability of decisions and results of AI (e.g. by storing training data) and ensuring that humans have the final say for important decisions.

The Commission is inviting people to submit comments on the proposals contained in the White Paper through a public consultation which is open until 19 May 2020. The White Paper can be viewed here.