Vulcan Insight

Parliament lays out its requirements for AI regulation

14 February 2020

On Wednesday the European Parliament adopted its position, by way of a non-binding resolution, on how to address the challenges arising from the rapid development of AI and automated decision-making (ADM) technologies. The Parliament’s opinion was delivered exactly one week before the European Commission will publish its approach to AI and ADM.

The Belgian Green MEP Petra de Sutter, author of the resolution on automated decision-making processes, urged the European Commission to guarantee “not only the free movement of AI enabled goods and services, but also the trust of the consumer in these new goods and services because the applications are numerous and encompass virtually all sectors of the internal market”.

MEPs agreed that four areas of future regulation need to be evaluated by the European Commission, namely: securing consumer choice, trust and welfare through transparency and human verification and oversight; updating product safety and liability rules for the digital age; extending existing rules to services incorporating automated decision-making processes; and protecting the quality and transparency in data governance to avoid the risk of biased AI.

The Parliament wants to protect consumer choice, trust and welfare and consider it imperative that automated decision-making can be checked and corrected with human oversight. This is important, the members say, in as much to improve efficiency and accuracy of services, as it is for the review structures and to verify decisions of  the automated systems that are being used to resolve disputes between consumers and traders. 

The outdated nature of the EU’s safety and liability rules is best exemplified by the fact that the EU’s product liability rulebook is 30 years old. The Parliament urged the Commission to propose updating the existing rules for the digital age by reassessing the definitions of ‘product’ and ‘damage’, as well as adapting the burden of proof. It is also encouraged that member states develop harmonised risk management strategies for AI.

The EU regulates services through a series of directives, including the e-Commerce Directive and the Professional Qualifications Directive. However, the rules were drafted with traditional services in mind. In light of the rapid development of AI and ADM, the Parliament have asked the Commission to propose extending the rules to services incorporating automated decision-making processes. The Commission is already tasked with proposing updates to the  e-Commerce Directive by the end of 2020.

The need for quality and transparency in data is necessary to drive innovation, but this requires good quality data. There have been high-profile instances of the use of biased data, including the decision by Amazon to scrap its AI recruiting tool that showed bias against women. In anticipation of potential future errors, the members support the creation of a review structure within business processes to remedy the mistakes. It is also urged that consumers have the possibility to seek human review and redress for automated decisions that are final and permanent.

The Commission will be considering the Parliament’s position as it finalises its proposal for a European approach to AI which is due next Wednesday 19 February. The resolution is, in effect, a list of minimum requirements needed to gain the support of the MEPs to the future regulation of AI.