Union Station Toronto corridor

AI Technology and Art: US Judge Finds that AI-Generated Art Cannot Be Copyright-Protected


1. Introduction

Is it really art if there is no human touch? At least for one American federal judge, the answer appears to be “No.” Artificial intelligence (“AI”) has recently become an accessible way of generating work that requires little to no human input. With software such as DALL·E 2 and Midjourney, the human ability to create has become filtered through analytical processes that often create stunning results. Dubbed “generative AI”, this technology uses raw data and produces new results based on patterns and machine learning.[1] However, as a recent American case found, these creations cannot be copyrighted.

This article provides insight on human authorship and how much is required for an artwork to deserve copyright protection. Next, it will examine how the US and Canada are currently responding to the rise of generative AI.

2. The US Cases

On August 18, 2023, Judge Beryl A. Howell found in Thaler v. Perlmutter (“Thaler”) that an AI-generated artwork could not be registered for a copyright.[2] The plaintiff, Stephen Thaler, had access to generative AI titled, the “Creativity Machine” and produced an image titled, “A Recent Entrance to Paradise.” After the US Copyright Office (“USCO”) denied Thaler’s application three times from 2019 to 2022, Judge Howell affirmed its rejection under judicial review.[3] You can view the image here.

In Thaler, Judge Howell wrote that “[h]uman authorship is a bedrock requirement of copyright”. They suggested that copyright was designed for humans for human-made work.[4] After all, a machine would not benefit from the conferment of exclusive rights that humans do. AI operates based on a specific degree of human input and follows orders to generate a result. Ownership of the AI will also not transfer ownership of the work itself to the human author.

Interestingly, Thaler was not the first case this year to treat AI and copyright. Earlier in February, the USCO revoked the rights of the comic book Zarya of the Dawn (“Zarya”) after the applicant failed to disclose that its images were AI-generated.[5] Later in March, the USCO issued a guidance stating that all copyright applications must disclose whether any part constitutes AI-generated content. A failure to disclose would result in a failure to register.

The current perception appears to be that AI is void of human qualities. Rather, both cases emphasize human authorship. In Thaler, Judge Howell found that it is the “sine qua non at the core of copyrightability”, whereas the USCO in Zarya found that works lacking a “minimum creative spark” are unprotected. It is inferred from both decisions that originality is not inherent in generative AI. Notably, AI cannot share the same qualities of independent creation and sufficient creativity that humans naturally possess, which are hallmarks of originality.[6]

Although AI can be used as a tool, both cases stress that the main driver of originality should be the artist themselves. Technology is useful at fostering human creativity, but it is not intended to replace it. However, both cases do not set a requisite standard of human input to determine what deserves protection. In Zarya, although the applicant argued that they used Photoshop to edit the AI-generated images, the USCO found them to be too minor. However, they wrote that “substantive edits” to AI art could potentially be eligible for copyright protection.[7]

3. Copyright Law in Canada

In Canada, the Copyright Act (“Act”) provides protection to individuals who are then solely capable of using their work for purposes such as producing, reproducing or selling it.[8] Although the Act does not provide a full definition, case law has attempted to define the meaning of an “original” work.

There are three ways to define “originality.” The first, dubbed the “Sweat of the Brow” camp, found that originality derives from one’s labour, skill and the expression of one’s thoughts, but not necessarily the uniqueness of ideas.[9] Second, the “Creativity” camp ushered in a low threshold of a “minimal degree of skill, judgment and labour in [a work’s] overall selection or arrangement”.[10] Lastly, the current leading case, CCH Canadian Ltd v. Law Society of Upper Canada, describes originality as such:

“…[A]n “original” work under the Copyright Act is one that originates from an author and is not copied from another work. (…) In addition, an original work must be the product of an author’s exercise of skill and judgment. The exercise of skill and judgment required to produce the work must not be so trivial that it could be characterized as a purely mechanical exercise.”[11]

Canada has yet to set out concrete steps to describe the relationship between originality and AI. However, one novel case involved an AI-generated painting titled Suryast, which was visually inspired by Vincent Van Gogh’s Starry Night. In December 2021, the Canadian Intellectual Property Office (“CIPO”) approved its copyright registration with the AI listed as a co-author, and the human applicant as the other.[12] Conversely, the USCO denied the applicant’s request for copyright protection in the US, where it is now awaiting review.[13] Although the USCO denied Suryast’s registration for being a “derivative work”, the CIPO appeared satisfied when a human co-author was also registered. Still, this raises new questions. One issue involves whether human co-authorship is enough to meet the requisite level of “originality”. Without set standards, this is difficult to answer. Furthermore, the degree of authorship is questionable depending on other factors, including the data from which the AI pulls or collects, whether its usage infringes any third-party rights, and how many prompts a user enters into the system.[14]

Today, no copyright laws in Canada address the topic of AI directly, let alone AI-generated works. One proposed statute tabled on June 16, 2022, called the Artificial Intelligence and Data Act (“AIDA”), is currently undergoing review as part of Bill C-27. AIDA is designed to provide a regulatory framework for AI that strikes a balance between fostering innovation and overall consumer protection. However, AIDA is notably silent on the issue of copyright.[15] You can read more about AIDA here.

4. AI Regulation in the US and Canada

No legal frameworks exist in the US and Canada that set out the requisite standard of human input in AI-generated art. Preliminary guides do exist, but they often neglect the issue of copyright.

In the US, the Department of Defense announced on August 10, 2023, the establishment of a generative AI task force, called “Task Force Lima”, which is intended to evaluate the pros and cons of this technology. This includes the safeguarding of national security and its potential benefits such as in commercial, health, warfighting, and policy affairs. However, it is unclear whether Task Force Lima will tackle the issue of copyright.[16] Nevertheless, the USCO’s stance on AI-generated works appears to be stricter than that of CIPO. As with the case of Suryast, the US belief appears to view any AI art to be unprotected and lacking in human authorship. Conversely, the Canadian approach seems rather lax. 

In Canada, Innovation, Science and Economic Development Canada has issued its own Code of Practice (“Code”) concerning generative AI. While it states that Parliament is intending to prioritize AI regulation once AIDA receives Royal Assent, the Code is silent on copyright.[17]

In March 2023, the “Canadian I.P. Voices” podcast by CIPO discussed AI art ownership. At issue were whether AI infringes the copyright of original works when pulling data to create new images, the ensuing liability, and the unclarity in whether a potential claim is viable against a piece of technology. It was proposed that a framework targeting AI and copyright would likely develop on a “country-by-country level”, although discussions between jurisdictions may help pave the way forward.[18] You can download the podcast here.

5. Takeaways

Currently, it appears that issues such as national security, transparency and accountability overshadow the ambiguity of whether AI-generated art is copyrightable.

Although Canadian courts have not yet rendered a decision on this issue, our American neighbours have developed persuasive arguments that can help us navigate relatively uncharted waters. While not binding in Canada, Thaler and Zarya are solid steps forward in staking a claim in what is copyrightable art. The registration of Suryast also fosters new debate in the amount of human authorship required for an original work.

Over time, it is likely that the courts will refine the interpretation of “originality” in light of generative AI. Admittedly, variables such as the choice of artistic medium, labour time and creativity, may make it difficult to set a definite threshold.

For more information about the legal implications of the use of AI technology, please contact Roland Hung of Torkin Manes’ Technology, Privacy & Data Management Group.

The author would like to acknowledge Torkin Manes’ Articling Student Herman Wong for his invaluable contribution in drafting this bulletin.

[1] Owen Hughes, “Generative AI Defined: How it Works, Benefits and Dangers” (7 Aug 2023) online: TechRepublic <https://www.techrepublic.com/article/what-is-generative-ai/>.

[2] Civil Action No. 22-1564 (BAH) [Thaler].

[3] Jon Brodkin, “US judge: Art created solely by artificial intelligence cannot be copyrighted” (8 Aug 2023) online: Ars Technica <arstechnica.com/tech-policy/2023/08/us-judge-art-created-solely-by-artificial-intelligence-cannot-be-copyrighted/>.

[4] Thaler, supra note 2 at 8.

[5] Re: Zarya of the Dawn (Registration #VAu001480196) (21 Feb 2023) online (pdf): United States Copyright Office <https://www.copyright.gov/docs/zarya-of-the-dawn.pdf> [Zarya]; Benj Edwards, “AI-generated comic artwork loses US Copyright protection” (23 Feb 2023) online: Ars Technica <https://arstechnica.com/information-technology/2023/02/us-copyright-office-withdraws-copyright-for-ai-generated-comic-artwork/>.

[6] Thaler, supra note 2 at 8; Zarya, supra note 5 at 3 (referencing Feist Publications Inc v. Rural Telephone Service Co., 499 US 340, 346 (1991)) & 10.

[7] Zarya, supra note 5 at 11–12.

[8] RSC, 1985, c. C-42 at s 3(1) & 5(1).

[9] MCM Consulting and Legal Services, “What is required for a work to be considered “original” for the purposes of the Copyright Act, R.S.C. 1985, c. C-42?” (2021) online: Westlaw Canada <https://nextcanada.westlaw.com/Document/Ie829c2dece511684e0440021280d7cce/View/FullText.html?transitionType=Default&contextData=(sc.Default)&VR=3.0&RS=cblt1.0>; Canadian Admiral Corp v Rediffusion Inc (1954), 20 CPR 75 (Can Ex Ct); Underwriters’ Survey Bureau Ltd v American Home Fire Assurance Co, [1939] Ex CR 296 (Can Ex Ct).

[10] MCM Consulting and Legal Services, supra note 9; Tele-Direct (Publications) Inc v American Business Information Inc, [1998] 2 FC 22 (Fed CA).

[11] MCM Consulting and Legal Services, supra note 9; 2004 SCC 13 [CCH Canadian Ltd].

[12] Canadian Copyright Database, “SURYAST” (1 Dec 2021) online: Government of Canada <www.ic.gc.ca/app/opic-cipo/cpyrghts/dtls.do?fileNum=1188619&type=1&lang=eng>.

[13] Sukanya Sarkar, “Exclusive: US Copyright Office refuses AI-assisted ‘derivative’ work” (3 May 2023) online: ManagingIP <https://www.managingip.com/article/2bm4rdot2clnuiwnrjd34/exclusive-us-copyright-office-refuses-ai-assisted-derivative-work>.

[14] Graham Hood, Alice Tseng & Patrick Roszell, “In-house counsel primer: Managing IP and compliance risks in artificial intelligence and a digital world” (11 Nov 2022) online: Lexology <https://www.lexology.com/library/detail.aspx?g=2f7d0f81-fc52-4960-8b20-7f2d2afcb9c0>.

[15] Innovation, Science and Economic Development Canada, “The Artificial Intelligence and Data Act (AIDA) – Companion document” (13 Mar 2023) online: Government of Canada<https://ised-isde.canada.ca/site/innovation-better-canada/en/artificial-intelligence-and-data-act-aida-companion-document#s11>.

[16] U.S. Department of Defense, “DOD Announces Establishment of Generative AI Task Force” (10 Aug 2023) online: U.S. Department of Defense <https://www.defense.gov/News/Releases/Release/Article/3489803/dod-announces-establishment-of-generative-ai-task-force/>.

[17] Innovation, Science and Economic Development Canada, “Canadian Guardrails for Generative AI – Code of Practice” (16 Aug 2023) online: Government of Canada <https://ised-isde.canada.ca/site/ised/en/consultation-development-canadian-code-practice-generative-artificial-intelligence-systems/canadian-guardrails-generative-ai-code-practice>.

[18] Canadian I.P. Voices, “Episode 26: Who owns AI-generated creations (and why you should care)” (1 Mar 2023) online: Government of Canada <https://ised-isde.canada.ca/site/canadian-intellectual-property-office/en/episode-26-who-owns-ai-generated-creations-and-why-you-should-care>.