Does AI have patent and copyright ownership?

 
Generative artificial intelligence (Gen AI) models have permeated both industry and popular culture with their ability to produce high-quality media, including text, code, images, audio, and video1. These models challenge our traditional understanding of creation and authorship, particularly in the context of securing intellectual property (IP) rights, leading legal and industry professionals to ask: how are IP authorities responding to Gen AI?

Questions of origin

Can AI be an inventor of a patent?

In Canadian patent law, an inventor is both i) the person who first conceives of the invention and ii) the person that sets the conception into practical shape2. In the age of Gen AI tools, it is also important to consider whether an inventor under Canada’s patent law must be human.

The issue of AI inventorship came before the Supreme Court of the United Kingdom (U.K. Court) in Thaler v. Comptroller-General of Patents, Designs and Trademarks ([2023] UKSC 49), which concerned two patent applications whose inventions were autonomously created by an AI program called DABUS. The U.K. Court ruled that inventorship was limited to natural persons, and so AI-devised material could not qualify as an invention. Therefore, since Thaler claimed that DABUS was autonomous, the U.K. Court held that Thaler did not have the right to apply for ownership of the patents himself. 

The fate of Thaler’s corresponding Canadian patent application has not yet been determined. When Thaler submitted similar patent applications in Canada, the Canadian Intellectual Property Office (CIPO) found that the designation of DABUS as an inventor was not compliant with the Patent Act, and Thaler responded by submitting an affidavit claiming that since he owned DABUS, he rightfully owned the inventions that it created. We are staying tuned for the Canadian perspective, as Thaler’s Canadian application is still pending before CIPO with the inventor listed as “unknown”3.

Can AI be an author of a copyrighted work?

The author of a copyrighted work is the person who expresses the work in its original form4. Traditionally, only humans have been listed as authors of registered works; however, in 2021 CIPO registered an AI program and a human as co-authors of an image called SURYAST5. RAGHAV, the AI program that produced the work, generates its outputs based on images provided by users. In this case, SURYAST was generated using two images: an urban photograph taken by the human creator, and Van Gogh’s famous painting, The Starry Night. The user also selected a numerical value representing the degree to which one image’s “style” transferred to the other.

Compared to patent applications, copyright registrations in Canada are subject to less extensive review and scrutiny by CIPO, which could be a factor relevant to the registration outcome of this work. Nevertheless, by registering SURYAST under co-authorship, CIPO appears to have recognized that the image generation process involved sufficient creative expression from both the user and the program. Whether this recognition could be challenged in court by an interested party, or CIPO would recognize a Gen AI program as the sole author of a registered work in Canada, remains unclear (for more on the intersection between artificial intelligence and intellectual property rights, read “Is it real or is it Memorex? Does AI threaten IP?”).

Are computer-generated works and inventions eligible for IP protection?

The Canadian perspective

With respect to patents, the Canadian government has yet to publish a formal position on the interplay of patent rights and computer-generated inventions.

With respect to copyright, however, the Canadian government published Consultation on Copyright in the Age of Generative Artificial Intelligence in 2023 to examine copyright policy considerations in the context of Gen AI.

Broadly, Canadian regulators aim to support AI innovation while maintaining incentives for traditional creators.

Broadly, Canadian regulators aim to support AI innovation while maintaining incentives for traditional creators. The report presents three possible approaches to copyright protection for computer-generated works: (i) limiting copyright protection to works created by humans; (ii) attributing authorship of AI-generated works to the person who facilitated their creation (i.e., the human user); or (iii) creating a new and unique set of rights for AI-generated works. The report raises additional complications that could influence which approach is ultimately taken. For example, a human author is necessary to quantify the duration of copyright protection, since rights expire 70 years after the author’s death. Implementing AI-exclusive authorship would require legislators to revise this system, while keeping rights duration fair between human and non-human authors (for more on Canada’s changing approach to AI regulation, read “What’s new with artificial regulation in Canada and abroad?”).

The American perspective

Earlier this year, the United States Patent and Trademark Office (USPTO) published Inventorship Guidance for AI-Assisted Inventions to clarify the extent to which AI programs can contribute to patentable inventions. These guidelines follow two USPTO decisions in 2020 which denied petitions from two related patent applications seeking to name an AI system as an inventor, on the basis that inventorship is limited to natural persons. Thus, solely AI-created inventions are categorically unpatentable. However, the guidance stipulates that AI-assisted inventions can still be patentable if a human makes significant contributions to the invention. In these cases, only the human would be listed as an inventor.

With respect to copyright, in 2023, the United States Copyright Office published a set of guidance entitled Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence. The guidance directs the Office to determine whether the AI’s role in the work’s creation was: (i) merely assistive to human authorship, or (ii) responsible for executing the traditional elements of authorship. In situation (i), the work is registerable under the human author; in situation (ii), the work lacks human authorship and is therefore unregistrable.

Most jurisdictions have yet to fully address the nuances at the intersection of AI and IP rights. Those interested in registering AI-generated inventions or works should monitor the regulatory landscape as it continues to develop.


  1. With grateful acknowledgement of the contributions made by Torys articling student, Rachael Gregoris.
  2. Apotex Inc. v. Wellcome Foundation Ltd. (2000), 10 C.P.R. (4th) 65 (F.C.A.) at para 30.
  3. New Brunswick Telephone Co. v. John Maryon International Ltd. (1982), 141 D.L.R. (3d) 193 (N.B. C.A.), leave to appeal refused (1982), 43 N.B.R. (2d) 468 (S.C.C.).
  4. Ankit Sahni. SURYAST. Canadian Copyright Database, 1188619, registered December 1, 2021, https://www.ic.gc.ca/app/opic-cipo/cpyrghts/dtls.do?fileNum=1188619&type=1&lang=eng.

This article was published as part of the Q4 2024 Torys Quarterly, “Machine capital: mapping AI risk”.

To discuss these issues, please contact the author(s).

This publication is a general discussion of certain legal and related developments and should not be relied upon as legal advice. If you require legal advice, we would be pleased to discuss the issues in this publication with you, in the context of your particular circumstances.

For permission to republish this or any other publication, contact Janelle Weed.

© 2024 by Torys LLP.

All rights reserved.
 

Subscribe and stay informed

Stay in the know. Get the latest commentary, updates and insights for business from Torys.

Subscribe Now