Artificial Intelligence: New Questions for Copyright Law

On December 1st, 2021, the Canadian Intellectual Property Office registered an artificial intelligence tool as a co-author for an artistic work. The piece is titled SURYAST, and this marks the first time a non-human has been attributed copyright authorship in Canada.

According to its Canadian registration (Reg. No. 1188619), SURYAST was authored by RAGHAV Artificial Intelligence Painting App as well as Ankit Sahni (the creator of the AI app). Like many humans, this AI app found inspiration in Vincent van Gogh’s portfolio. However, unlike humans, the app’s inspiration took the form of being fed a base dataset derived from van Gogh’s Starry Night.

Practically speaking, what does all of this actually mean? Having copyright registered in Canada creates a presumption that copyright exists in the work and that the registrant is the lawful owner. This presumption shifts the burden of proving otherwise onto any party challenging these facts. In the case of SURYAST, there are several practical issues that have been observed in relation to this presumption. For example, how would an AI access its rights as author and/or owner of a work? And, how could it defend its registration in court? However, there are potentially even more concerning questions out there.

Observers have generally focused on this AI being named as a joint author. However, there is something potentially even more notable on SURYAST’s registration: the AI is not named as an owner of the work – Ankit Sahni (the human) is the registered owner. In the absence of an employment relationship, the author of a work is typically the first owner of the work as well. How did Ankit Sahni become the owner of SURYAST? Did the AI, as joint author, somehow agree to assign ownership to Ankit? How can an AI have the capacity to make this assignment? Surely this AI is not so advanced as to possess and exercise the autonomy necessary to contract with others. Can the AI be deemed to be an “employee” of Ankit?

All of these questions fall under a larger ongoing debate around reform of the Copyright Act.[1] A central question in this debate is how to avoid Canadian Copyright law hampering innovation. For example, how to approach the use of copyright-protected works when training AI (i.e. feeding it a dataset of copyrighted works).  Does the “use” of these works in this manner constitute copyright infringement?

For now, it is unclear how Canada will treat the relationship between AI and copyright law. It remains a constant policy challenge in Copyright law to balance the proprietary rights of authors and employers with the benefit of making new innovations broadly available.

[1] https://www.ourcommons.ca/Committees/en/INDU/StudyActivity?studyActivityId=9897131