Online hate speech and intermediary liability in the age of algorithmic moderation

This research aims to investigate the impact of liability-enhancing legal strategies in the context of the governance of online hate speech. Indeed, the increased reliance of the law on the role of private platforms for the purposes of moderating and removing hate speech deeply affects constitutional principles and individual fundamental rights. For instance, the enhancement of intermediary liability and responsibilities can con-tribute to the phenomenon of the over-removal of user content, with little regard to basic constitutional guarantees. Furthermore, research has shown that the ever-increasing use of automated systems for hate speech moderation gives rise to a whole new set of challenges and issues related to the concrete risk of errors and biased results, leading to a disproportionate removal of content produced by minority, vulnerable, or discriminated groups of people. After dealing with the question concerning the ra-tionale(s) of hate speech regulation and arguing for an increased role for the principle of substantive equality in this regard, this work investigates the developing trends con-cerning the imposition of forms of intermediary liability with respect to the spread of hate speech content across the Internet, keeping a close eye on the evolving European framework. In doing so, this work also explores the relationship between platforms’ content moderation practices and the promotion of fundamental rights and values – in-cluding the principle of substantive equality – especially in the light of the ever-increasing use of artificial intelligence systems for the detection and removal of hate speech. In the context of the European Union, it is held that such reflections are of ut-most importance particularly following the adoption of the Digital Services Act. In this respect, the work argues for the need for a renewed code of conduct on hate speech, with a view to further protecting constitutional values and the rights of users.