Article 38 Recommender systems

The obligations on assessment and mitigation of risks should trigger, on a case-by-case basis, the need for providers of very large online platforms and of very large online search engines to assess and, where necessary, adjust the design of their recommender systems, for example by taking measures to prevent or minimise biases that lead to the discrimination of persons in vulnerable situations, in particular where such adjustment is in accordance with data protection law and when the information is personalised on the basis of special categories of personal data referred to in Article 9 of the Regulation (EU) 2016/679. In addition, and complementing the transparency obligations applicable to online platforms as regards their recommender systems, providers of very large online platforms and of very large online search engines should consistently ensure that recipients of their service enjoy alternative options which are not based on profiling, within the meaning of Regulation (EU) 2016/679, for the main parameters of their recommender systems. Such choices should be directly accessible from the online interface where the recommendations are presented.