Registration

A stolen body: When AI steals our dance moves

Literature, visual art, music, theatre. Most artistic professions have already entered the fight against content generated by artificial intelligence (AI). Photography competitions and stock photo databases are setting rules to counter AI photos out of fear of losing market value. Generators like MidJourney train on millions of existing pieces, often without author consent. AI is able to compose music in the style of well- known artists, for example, the fake song "by" the rapper Drake, which sparked a discussion as to wether it was eligible for a Grammy (Grow 2023). Writers are coming out strongly against projects which generate books based on their style. George R. R Martin sued OpenAI in 2023 (Associated Press 2023). Just a couple of months later, actor Scarlett Johansson took legal action against the developers of Lisa AI, which used her likeness and voice for their advert without her permission (Parkel 2023). And what about the world of dance? In recent years, AI has been used in a host of experimental projects, bringing innovative forms to the stage and addressing audiences in new ways, but, as with other professions, here we must also question from whom did AI learn to dance and under what conditions?

Created by Leonardo.AI
Created by Leonardo.AI

Copyright in choreography in the age of AI
Dancers and choreographers have not yet made a lot of noise in this regard. In their case, we are not usually talking about millions in royalties and valuable licensing agreements. When there are disputes in relation to unauthorised use of movement, it is largely connected to the gaming industry. The issue with ownership of movement is all the more complicated in the digital age, since dance historically relied on interpretative tradition rather than rigid copyright. The latter became a live issue in dance around the turn of the 20th century.

The turning point came with the development of copyright legislation itself, but dance was far behind literature and music in this regard. One of the first people to recognise choreography as an artwork entitled to copyright protection, was Michail Fokin (1880–1942). He repeatedly fought against the theft of his ballet ensemble choreography and systematically stood up for the idea that choreography is an original artwork that should not be arbitrarily copied and restaged without the author’s consent. Around the same time, the first disputes of this kind began to appear, in many ways reminiscent of today’s copyright cases – especially in terms of whether it is possible to own a movement form, structure, or composition. So, choreography copyright was legally established in most countries throughout the 20th century (e.g. in the USA, choreography officially became a protected authorial work with the introduction of The Copyright Act of 1976).

At the same time, there are, of course, many artists whose legacy is carefully guarded (the most well-known being Kenneth MacMillan, George Balanchine, Frederick Ashton, John Cranko, among others), and even though there have been initiatives to draw greater attention to this problem and authors have been published on the subject.1 Unlike works of literature or visual art, where copyright operates as a decent defence against unrestrained copying, dance continues to enjoy very little legal protection. It is hard to categorise movement as “intellectual property”, so AI does not need any permission to appropriate it.

Today, algorithms regularly analyse choreographic patterns without the author ever agreeing to be part of the digital archive. Yet, no- one is currently able to stop, for example, MacMillian’s well-guarded production of Romeo and Juliet (of which there are hundreds of recordings on YouTube alone) being used for machine learning and then even copied in some form by AI. Finding the line between a step, which is hard to copyright, and a sequence of steps, which may be considered original choreography, has been a great challenge for the dance community since the origins of the copyright discussion, but today, when it is possible to use AI to read and generate virtually in real time, this problem becomes even more complicated, and patently more urgent.

Contribute to content creation

Interested in the full article?

Purchase the individual text or the entire issue of Dance Context Journal.

What you’ll find in the Journal:

  • Season Summaries in Contemporary Dance, Ballet and Nonverbal Theatre
  • Popular science articles on current topic 
  • Scholarly studies on current topic

Buy article         Buy Dance Context Journal