A European solution for Text and Data Mining in the development of creative Artificial Intelligence – With a specific focus on articles 3 and 4 of the Digital Single Market Directive

Publicerad i Stockholm IP Law Review 2021 #2, december 2021 s. 18–33


In today’s data-driven society, Text and Data Mining (TDM) has become an essential tool in managing Big Data in its different sizes and forms, which is also an inherent part of Artificial Intelligence (AI) research. TDM techniques highly depend on datasets derived from TDM to self-learn and to make autonomous decisions. Through the lens of copyright and related rights, TDM may be used to train AI for the purpose of AI-driven creativity, where AI has already helped in completing paintings, composing music and producing movie trailers. However, since TDM typically involves the acts of copying and/or extracting of works and other subject-matter protectable by copyright and related rights, legal restrictions under the EU acquis might be in place.

In this regard, the importance of TDM has been acknowledged by the EU Legislators, which introduced two mandatory exceptions for TDM contained in articles 3 and 4 of the Directive 2019/790 on copyright and related rights in the Digital Single Market (DSM Directive). The present article analyses the relationship between TDM and AI-driven creativity by, firstly, explaining the notion of TDM and common technical steps within its process. Secondly, it examines the copyright and related rights issues regarding TDM and possible pre-existing exceptions and limitations under the EU acquis that might be applicable. Lastly, this article critically analyses the mandatory TDM provisions under the DSM Directive and concludes that these still contain shortcomings that may significantly restrict the possibility to undertake unlicensed TDM for AI creative purposes within the EU.