The Copyright Act Needs to Address AI

Over the past decade, the capabilities of technology and AI have rapidly increased. While creativity used to be something that was solely attributed to the human experience, AI can now be used to create things like artwork and literature. A simple Google search on AI-created art leads you to several apps promising to “transfer words into art” or to “recreate some of the world’s most legendary paintings.” Simple algorithms have been used to write new chapters of famous books like Harry Potter, or television episodes for sitcoms like Modern Family. But while AI capabilities have changed so drastically over the years, the copyright laws around them have not. With all of these AI-creations entering the world, it brings up the question: who owns them? 

            The discourse over AI copyright is something that has been rising over the past few years. For the sake of narrowing down the scope, this paper is going to focus specifically on Canadian Copyright, and what the Copyright Act dictates should be the case when it comes to work created by AI–and whether its current rules are appropriate. 

            The Copyright Act of Canada was last updated in 2012 and this modernization of the act was actually instated in order to catch up with the rapidly increasing technology and the popularisation of the internet. Barry Sookman, an expert in evolving technology at the time, stated that the modernization would: “update the rights and protections of copyright owners to better address the challenges and opportunities of the Internet, so as to be in line with international standards; clarify Internet service providers’ liability and make the enabling of online copyright infringement itself an infringement of copyright; permit businesses, educators and libraries to make greater use of copyright material in digital form; [and] ensure that it remains technologically neutral (Guly).”

            But at the time of this update, AI technology was still in its infancy stages. 2012 boasted technological advancements like the PS Vita gaming system, the Android launch of the Instagram app, and the introduction of the Kindle Paperwhite e-reader (Engadget), which meant that lawmakers were not considering AI-created works with these new copyright laws, but rather how to protect human creators when there were new platforms arising for their work to be presented on. It’s doubtful that any of them saw technology like “Bot Dylan,” an AI folk musician that writes songs based on an algorithm that holds samples of over 20,000 songs (Gray), coming.

            In the Canadian Copyright Act, a copyrighted work necessitates an author. The author of the work is the first owner of the copyright, and the duration of the copyright lasts until 50 years past the date of their death. Due to this wording, it can be assumed that authors must be alive—thus excluding AI from consideration. However, a recent case in Canadian copyright challenges this.

            In December 2021, Ankit Sahni created a painting with the use of an AI program called RAGHAV. The copyright claim for the painting Suryast, which was inspired by the famous Starrey-Night painting by Vincent Van Gogh (, was filed with Sahni and the program as co-authors, making RAGHAV the first AI program to ever hold a copyright claim in Canadian history (Holloway). In this case, the AI was only a co-author—meaning that there is still a person there to hold the copyright if there were ever challenges with their intellectual property. But it brings up questions on what would happen if an AI could independently hold a claim for their creations.

            The first problem that arises with this is the question of ownership. Even if an AI creates a work independently, human intervention was required to create the AI. The AI’s creator may lay claim to the AI’s coding and algorithm, whereas a second person may have chosen the data to be used in those algorithms to bring about the end result. Even investors could argue for their part in funding the project, as well as the companies that created the parts that were used in the AI computer. The sheer number of people involved to bring an AI “to life” is so large that it can be difficult to say that an AI originally created the work on its own.

            A second argument that can be made on whether the works are truly transformative enough in nature to be copyrighted. An AI is not able to generate ideas from nowhere—they do not have the processes for true creativity and originality. Using the “Bot Dylan” example from before, the bot was not able to simply write folk music without a guide. It analyzed those 20,000+ songs in its algorithm in order to create something from it. But depending on how much it took from those already-existing songs, it could be argued that it did not create anything new and that the original songwriters still owned the credit.

            What would it mean for an AI to hold the power of authorship? The guidelines of the act would have to change entirely. After all, machines cannot hold residency—which is a defining factor of Canada’s current copyright laws. And if an AI can be infringed upon, it also must be held responsible for infringing. How would a criminal punishment against a machine be enacted? In the same way that machines wouldn’t care about an infringement on their rights, a machine also wouldn’t care about consequences given to them for their crimes.

            When the facts are laid out like this, it becomes clear that AI should not be given the right to hold copyright. But this is a question that has been heavily debated over the past few years and is a rising concern in Canadian copyright. In fact, just last year the government launched a public consultation to gather insight on the public’s thoughts on several issues revolving around technology and copyright: one of which was about “authorship and ownership of AI-created works; infringement and liability regarding AI (Guly).”

            In the consultation papers that the government released last June, it appears that they are running into the same problems defining AI authorship as I outlined in this paper. While some consults suggested that AI-created works remain in the public domain, some argued that a distinction between AI generating work with human intervention and without needed to be made. But from there only comes more questions: how do you define human intervention? Does the fact that Bot Dylan—even if he runs on his own without a human helping with the decision making—uses songs written by humans count as human intervention? In that case, will there ever truly be an AI that can create something entirely without human help?

            Right now, all the Copyright Act leaves is questions. A better-defined law around AI-created works is definitely in need. This inquiry is hopefully only the first step—but hopefully law-makers can make a decision before the technology evolves even more.

Bibliography

Engadget. “2012 Year in tech: A timeline.” Engadget, 28 Dec. 2012, https://www.engadget.com/2012-12-28-2012-year-in-tech-a-timeline.html

Government of Canada. “A guide to copyright.” Government of Canada, 2022, https://www.ic.gc.ca/eic/site/cipointernet-internetopic.nsf/eng/h_wr02281.html?Open&wt_src=cipo-cpyrght-main&wt_cxt=learn#additionalConsiderations

Guly, Christopher. “Canada Is Gathering Public Input on Copyright Implications of AI, Internet of Things.” Center for International Governance Innovation, Aug. 2021, https://www.cigionline.org/articles/canada-is-gathering-public-input-on-copyright-implications-of-ai-internet-of-things/

Holloway, Jim. “Welcome to the machine? Canadian copyright protection of AI-generated painting.” Baker Mckenzie, 22 Feb. 2022, https://viewpoints.bakermckenzie.com/post/102hjfi/welcome-to-the-machine-canadian-copyright-protection-of-ai-generated-painting

Mederios et. al. “IP monitor: Copyright protection for AI-created work?” Norton Rose Fulbright, Mar. 2022, https://www.nortonrosefulbright.com/en-ca/knowledge/publications/68947aaf/copyright-protection-for-ai-created-work

Leave a Reply

Your email address will not be published.