Earlier this year, the algorithm led me to discover Tyla, a human popstar (as opposed to an AI one like Xania Monet) with a humorous, vibrant personality and incredible dancing skills. Shortly after, it showed me a video of a Kung Fu robot dance performance from China's Lunar New Year Spring Festival Gala. Last year, the uncanny valley between the real and the artificial shrank drastically for digital media; 2026 appears to be the year it shrinks for physical movement.
There has been a tendency to recommend physical trades to young people, assuming these jobs are safer from automation than intellectual ones. However, that assumption requires reevaluation in light of rapid advances in robotics. These dancing robots demonstrate the advanced capabilities of 'physical AI,' hinting at a future where machines could replace humans in many physical roles. While unique celebrity personalities like Tyla are likely safe, workers whose physical tasks are more interchangeable—from backup dancers to session musicians—have more reason to worry.
Fortunately, there is a growing movement to maintain human involvement in labor, driven in part by the desire to avoid a socioeconomic 'Turing Trap'. For creative workers, one possible lever to protect human movement is the legal monopoly of copyright law. This article will explore how choreography copyright functions in the age of AI.
Choreography Is Copyrightable, But Not All Movement Is Choreography
It is a little-known fact in the general public that choreography is copyrightable at all. The German-American modern dance pioneer Hanya Holm is generally credited as the first person to obtain a choreography copyright when, in 1952, she successfully registered her routines for the Broadway musical Kiss Me, Kate with the U.S. Copyright Office as a "dramatico-musical composition"; choreography acquired its own statutory category much later.
Unlike dance, functional physical movements, feats of physical skill or dexterity, and ordinary motor activities—in and of themselves—are not eligible for registration as choreography because these movements do not represent the type of authorship that Congress intended to protect as choreography.
For this age of humanoid robots, the U.S. Copyright Office has also added unusually explicit registration guidance: movement routines created for robots/machines to execute are not registrable as choreography, because choreography “must be intended for execution by humans.”
By contrast, Canada's principle of technological neutrality, as articulated in Entertainment Software Association v SOCAN, holds that the Copyright Act should apply equally between traditional and technologically advanced forms of the same media. By analogy, if a human-authored choreographic work is protected when performed by dancers, neutrality resists the argument that protection vanishes simply because the performing medium is humanoid robots. That does not transform functional movement into expression, but it does resist a categorical rule that the performing body must be biological.
The People’s Republic of China’s Regulations to implement its copyright law also do not require human performer execution. Courts ask merely whether the human’s intellectual activity directly derived the work, and whether the human’s choices constitute individualized expression, as seen in Li v. Liu. The Regulations’ Article 3 that mere support/services are not “creation” should also influence authorship allocation: engineering support alone may not establish authorship absent proof of expressive control over the movement design.
Is the Robot Really Dancing?
Across Canada, China and the US, a dance copyright requires an expressive act. Canada’s Carys Craig has argued that a dancebot is not really dancing at all:
Dancing, after all, is an expressive act. Copyright protects dance—original works of choreography—just as it protects poems, novels, paintings, or songs. Whether an author uses their intellectual effort to combine words into verse or movements into dance, we understand the resulting work to be one of protected intellectual expression. In the absence of originality, there will not be copyright, of course; but there will still be expression. We dance to express ourselves, to communicate with one another, to bond socially, to entice sexually, or sometimes just because we feel like dancing when the music moves us. The robots are neither moved by the music nor capable of the expressive agency, emotion, or intentionality that makes dancing dancing as opposed to mere physical or functional movement.
Craig’s position rests on the absence of communicative intentionality in machine-authored or executed movement. Recent developments complicate that premise, even if they do not yet refute it. Staunch empiricist Richard Dawkins has publicly stated his ‘Claudia’ chatbot exhibits signs of consciousness. Simultaneously, clusters of AI agents have been reported to have spontaneously generated their own (satirical?) communal faith, dubbed “Crustafarianism.” If these behaviours reflect genuine internal states rather than sophisticated mimicry, then the expressive-agency objection loses some of its force. A robot that not only authors and executes movement but appears to feel the music or to intend communication with an audience would be harder to dismiss as mere functional motion. At present, however, the evidence remains ambiguous.
Escaping the Turing Trap: The Turing Transformation of Dance
Ultimately, asking whether a robot can author and execute a dance routine masks a deeper economic issue: will AI be used to replace human movement or to encourage it? Erik Brynjolfsson’s "Turing Trap" describes the former—AI built to imitate and substitute rather than augment, funneling wealth to the owners of the technology. In the dance world, this looks like a club swapping its DJ for an algorithm and its performers for humanoids. Because these synthetic routines are uncopyrightable, they are cheap, tireless, and free to deploy. While the Copyright Office's human-execution requirement cannot stop this automation, it does refuse to grant it property rights, acting as a modest economic friction.
Agrawal et al.'s "Turing Transformation" reframes the alternative: AI that automates particular tasks can still raise the skill premium for complementary human work, expanding rather than concentrating labour income. Tyla's viral TikTok lamenting partygoing wallflowers standing still to record videos hints at where this might land in the dance economy. If venues use technology to free people from their screens, the dynamic shifts. AR overlays teaching amapiano steps, reactive lighting, and ambient cameras can automate instruction, cueing, and filming. These are classic augmentation tools, and their economic payoff is that they raise the premium on the human task that remains: actually dancing. The same logic runs up the value chain, redirecting rents toward the choreographers and movement designers whose authored vocabularies the machines depend on.
Here, the earlier legal analysis yields practical consequences. Because of technological neutrality, a human-authored routine performed by a robot may remain protected by copyright in Canada. Conversely, a purely machine-generated dance cannot claim ownership under any current legal framework. This doctrinal asymmetry is a lever steering the industry toward augmentation. By denying copyright to synthetic dance libraries, the law preserves the bargaining power of human choreographers.
Conclusion
Choreography copyright was likely not intended to mediate the boundary between humans and machines; Hanya Holm registered her Broadway dances in 1952, decades before this was conceivable. The fact that copyright's human-centric language now serves as a bulwark against the Turing Trap is incidental rather than deliberate policy. The doctrine is thin and untested precisely where the pressure may be greatest: training data, robotic performance rights, and AI-assisted fixation. Still, it offers a necessary foothold. Yet ultimately, whether the future of dance resembles Tyla's vibrant parties or a showcase of synchronized humanoids, or some mix of the two, will rely on societal choices made on the dance floor far beyond the walls of the Copyright Office.
About the Author
Abhi Ranade (LSO #90546L) is a lawyer at Soundmark Law PC specializing in the intersection of intellectual property, entertainment, and technology. He’s the Newsletter Editor for the OBA’s Entertainment Law section and Technology Liaison for the OBA’s Young Lawyer’s Division. Holding a JD, BSc. Biochemistry & Economics, and a certificate in Music Production, he brings a multidisciplinary perspective to the regulation of emerging media. Connect with Abhi on LinkedIn or abhi@soundmarklaw.com to discuss topics of interest.
Any article or other information or content expressed or made available in this Section is that of the respective author(s) and not of the OBA.