Algorithmic fashion and copyright: the regulation of text and data mining to detect and anticipate future trends
This is a brief excerpt of an upcoming research paper wrote by Prof. Eleonora Rosati, Associate Professor in Intellectual Property Law at the University of Southampton and Of Counsel at Bird&Bird, for Ius in itinere’s focus on Fashion Law coordinated by Nicola Lanna, Fashion Law and Influencer Marketing expert Lawyer.
‘Algorithmic fashion’ is a phrase used to indicate the use of algorithms to anticipate and, thus, seize the fashion trends of the future. This is not an isolated phenomenon: it is becoming commonplace also in the music and film industries.
But how are future trends detected and exploited?
Social media and the content shared by users therein plays a significant role. For instance, in the context of an anthropological study conducted at Cornell University, text and data mining (TDM) techniques have allowed a group of researchers to mine 100 million photographs made available on Instagram and devise patterns on how clothing styles vary around the world, and tackle the frequency of use of certain garments and colours. By training a machine-learning algorithm, the researchers were able to identify a set of visual themes and study how these would vary by time and place, and also identify the preference for certain colours.
All this raises a number of questions, including questions related to copyright: can it be that, by copying the photos shared by Instagram users, the researchers did copyright-restricted acts and, potentially, infringed copyright?
TDM may be performed in a variety of fields and for different purposes. Among other things, TDM techniques may be used to ‘train’ Artificial Intelligence (AI), also for the purpose of AI-driven creativity. In this context, copyright restrictions might be in place, even if copies made of pre-existing content are only used internally and are instrument to the creation of something else.
Recently, in the context of the Directive on copyright in the Digital Single Market (Directive 2019/790), the EU legislature introduced two new mandatory exceptions for TDM.
Although less developed than in other jurisdictions, TDM techniques display a clear potential also for European businesses. In this sense, a mandatory exception allowing TDM also for subjects other than research organizations and cultural heritage institutions and for purposes other than scientific research – as is the exception in Article 4 of the Directive – is welcome. However, the freedom left to rightholders to ‘reserve’ the right to carry on reproductions of their subject matter, including for the purposes of TDM, might in practice hinder the application of the exception in Article 4 and even defeat its purpose altogether.
In addition, by framing TDM within the scope of copyright protection, EU legislature has clarified once and for all that the undertaking of TDM activities outside the scope of a licensing agreement or qualification for the protection offered under available exceptions would expose one to potential liability for copyright infringement, irrespective of whether the process of copying (if any) is intermediate and finalized at extracting what copyright law does not protect, that is facts, data, and information. In practice, this might have a negative impact on AI creativity.
Also fashion businesses based in Europe should take notice.
The newly adopted TDM exceptions, notably the one in Article 4, could enhance the development of algorithmic fashion, though it might be unlikely that mining activities could be performed in each and every instance without a licence.
 K. Matzen et al, “StreetStyle: Exploring World-Wide Clothing Styles from Millions of Photos”, (2017).