Encourage museums to use digitized contents for machine-learning purposes, but do not grant copyrights to machine-created digital content

A – Issue at stake

  • Museums hold an important amount of data, including various types of metadata pertaining to the physical and born-digital works in their collections which, if properly structured, allows museums to analyze collections, objects, and creators in novel ways.1 However, for data to be fully exploitable and exploited, museums must put machine-learning processes in place, e.g. allow artificial intelligence (AI) to mine through their digitized collections to “learn” from the data, process it and produce results; indeed, “AI [is] an essential tool for museums managing the massive scale of data in the 21st century [and] may be the only feasible way to make meaning of archives at this scale.”2 This allows museums to notably “unlock the potential of digital image collections by tagging, sorting, and drawing connections within and between museum databases”3, keep track of newly generated data as archives grow, and help identify fakes and forgeries.4 AI may even act as a curator and put together new collections with the help of algorithms based on image metadata, creating unexpected and surprising combinations that would not be perceived by a human mind.5
  • Moreover, works fully created by machines and algorithms are now being sold by world-leading auction houses,6 featured in exhibitions7 and even dubbed as more novel and appealing than “real” works.8 It is only a matter of time until such works integrate museum collections.
  • This advent of AI in museum activities however raises important copyright questions that have not yet been clearly answered. For instance, there is still uncertainty as to whether copyright limitations and exceptions allow the use of existing digital content by machines for learning purposes (although a majority of authors seem to believe that they do).9 The copyright status of AI-created works is even more unclear10 and multiple issues reaching far beyond the realm of copyright will need to be taken into consideration before we reach a final answer.11
  • This situation may deter museums from exploiting AI to its full potential and prevents them from fulfilling their public interest mission in accordance with the 21st century technological developments.

B – Clarifications

  • Museums should be encouraged to use AI to fulfil their public interest mission.
  • Policymakers and legislators should take a clearer stance regarding the use of existing digital content by machines for learning purposes, by specifying in legislation or official policies that it does not constitute an infraction to copyright law.
  • Furthermore, priority should be given to research pertaining to the copyright status of AI-created works. Policymakers and legislators should specify that those works are out-of-copyright, which will facilitate their dissemination and re-use by museums and the public.