Is the output of the generative AI system protected by intellectual property rights?

Case law has discussed the need for independent intellectual human effort before the originality requirement (and connection to Australia requirement) is met.² This, of course, implies that an author must be human. That does not mean that computer assistance is prohibited, but where human input is negligible (as it likely would be in generative AI developed materials), those Outputs would not attract copyright protection absent significant User input to direct the creation process.³

This outcome has implications for ownership in a number of significant ways:

• Vesting provisions in the Copyright Act 1968 will have no effect without an author, since the statutory presumption is that the author of a work is its owner subject to any previous agreement contemplating assignment or creation in the course of employment.
• Even where there is a Provider/Deployer/User agreement, neither could assign title to something they do not have title in themselves because there is no copyrightable material being created.
• As a consequence, AI-generated Outputs are likely to be public domain.

Patents: Whether AI could be considered an inventor within the meaning ascribed in the Patents Act was settled in the recently decided Thaler case.⁴  The Federal Court of Australia:

• Surveyed the relevant test for which a person would be entitled to be named as inventor.
• Did not say that ‘person’ in relevant cases, in the Act and its regulations meant that it had to definitively be a natural person.
• Did say, however, that the grant of a patent was a reward for the ingenuity that sprung out of the ‘mind of a natural person or persons.’

Copyright: Was human skill and judgment used to generate the output? In Canada, copyright applies to a work where the author was, at the date of the creation of the work, a citizen or subject of, or a person ordinarily resident in, Canada or some other treaty country.⁵

Copyright protection for ‘data’ will depend on the nature of the data and how the data was created. Compilations of data are eligible for copyright so long as skill and judgment was employed in the selection or arrangement of the data.⁶  Copyright can protect computer-generated data provided that skill and judgment was used to generate the data.⁷

In Canada, AI-generated works may be eligible for copyright protection. Current requirements for copyright protection include originality, fixation and the exercise of skill and judgment. These requirements can be understood to mean that content created by generative AI is not protected unless an original work is created with the sufficient addition of human skill and judgment.

For example, in December 2021, the Canadian Intellectual Property Office (CIPO) registered a copyright for a painting entitled ‘Suryast,’ with two co-authors: Mr. Ankit Sahni and RAGHAV Artificial Intelligence Painting App, making ‘Suryast’ the first Canadian copyright with an AI author.⁸

As another example, in April 2023, CIPO granted a copyright registration for an image and a poem entitled ‘Sunset Serenity, being an image and poem about sunset at an Ontario lake created entirely by AI programs DALL-E2 (Image) and ChatGPT (Poem) on the basis of prompts demonstrating minimal skill and judgment on the part of the human claiming copyright.’ The copyright registration only lists a human as an author and owner, despite the title referring to a generative AI system.

It remains unclear if CIPO would register copyright based on an application that has no human author listed. While there is no explicit definition of the term “author,” Canadian case law has traditionally held that an author ‘must be a natural person’ ‘who exercises skill and judgment’ in creating the work, but listing a human co-author for an AI-created work seems to meet that requirement. Further, the above-noted copyright registrations may be later challenged in court.

Patents: In Canada, the Patent Act does not define ‘inventor,’ much like the Copyright Act does not define ‘author.’

Recently, Dr. Stephen Thaler filed patent applications in various patent offices worldwide for a patent with the sole inventor listed as ‘Device for Autonomous Bootstrapping of Unified Sentience,’ or DABUS, an AI machine created by Dr. Thaler and designed to invent. In December 2021, CIPO issued a non-compliance notice for the DABUS patent application because ‘the inventor is a machine and it does not appear possible for a machine to have rights under Canadian law or to transfer those rights to a human.’

However, the CIPO notice also stated that the applicant, Dr. Thaler, may attempt to comply by submitting a statement on behalf of the AI machine identifying himself as the legal representative of the machine.

In July 2022, the applicant filed a letter submitting that the applicant is entitled to file a patent application through his ownership of DABUS. The CIPO examination of the DABUS patent application is still ongoing at the date of publication. This case raises a question of whether this patent application would have overcome this initial threshold determination if Dr. Thaler had been named as a co-inventor.

Copyright: While the Copyright Law of the People’s Republic of China does not explicitly mention AI-generated works, it provides a basis for their protection within the existing legal framework. The Copyright Law and relevant court cases indicate that AI-generated content can potentially meet the criteria for copyright protection under certain conditions.

The Copyright Law:

• Defines ‘work’ as original intellectual creations in various domains that can be reproduced in a tangible form.
• Does not explicitly exclude the possibility of AI-generated content being protected by copyright. While the law does not provide a clear definition of ‘originality,’ the fact that AI-generated content can meet the criteria of being independently completed (and be different from prior works) suggests the potential for copyright protection.

However, there are considerations related to human involvement in the creative process. Whether a work generated by AI can constitute an ‘original’ work within the meaning of Copyright Law will largely depend on whether the work has conveyed the unique expression of thoughts and emotions of a human – for example, the Deployer or User, behind the AI.

In Beijing Feilin Law Firm v. Baidu Corporation,⁹  the Beijing Internet Court and the Beijing Intellectual Property Court:

• Held that copyright did not exist in a chart and corresponding figures generated by the plaintiff Beijing Feilin Law Firm using an AI tool powered by Wolters Kluwer. The chart and figures generated by the AI tool were said to lack the thoughts and emotional reflection of either the developer or the User of the AI tool.
• Found that copyright subsisted in the remaining parts of the work at issue (a report analyzing particular judicial decisions). The analysis part of the work reflected necessary originality of the creator.

When an AI-generated work is the ‘original work’ under copyright law, the courts could find it to be the work of the Deployer or the User using the AI. The courts in China still consider AI to be a tool. For example, the Guangdong Court in Tencent v. Shanghai Yinxun Technology Ltd.:¹⁰

• Took the view that an article produced by AI reflects the personalized arrangements and choices of the plaintiff and has a certain degree of originality.
• Held that the work in question was created by an AI system and was a literary work protected under copyright law. The plaintiff, Tencent, as a legal entity was found to be the author of the work due to their supervision and organization of its creation.

Patents: The Chinese Patent Law and Patent Examination Guide make clear that, in China, the inventor of an invention should be a natural person,¹¹ and the patentee of an invention can be an individual or an organization.¹²

Accordingly, AI itself cannot be the patentee of an invention. If the User has made sufficient contributions to the invention, they may be named as the patentee or inventor of a patent.

Author’s rights (copyright): Author’s rights protection is granted to works meeting the criteria of originality, defined by case law as the author’s ‘personality imprint.’ The qualification of author is not only granted to individuals but also (subject to conditions laid out under the French Intellectual Property Code for collective works) to legal entities.

French law does not explicitly address AI-generated works. Therefore, ownership might depend on factors such as the level of human involvement in creating the AI system or curating its Output. For instance, if the AI system itself generates content without direct human involvement, the legal concept of ‘authorship’ might not easily apply, and it would be extremely difficult to meet the criteria of originality, given that it is tied up with ‘someone’s personality.’

Rights in data: No property right or intellectual property rights subsist in data per se under French law. Where the Output is a database, this can be protected under the sui generis database right. The French Intellectual Property Code protects a database in which a substantial investment has been made (either financial or in terms of time and effort) to obtain, control or present the contents of the database. The rightsholder is the (natural or legal) person who bears the risk of the investment to be made for the database and who spent effort in obtaining, verifying or presenting the data; such person is given the exclusive right to grant permission for the retrieval or reuse of a substantial part of the contents of the database (systematic retrieval of small parts of the database is also not allowed).

It remains unclear the extent to which a Provider or User of a generative AI system is likely to benefit from this right in practice. The threshold for demonstrating substantial effort in obtaining, verifying or presenting the data is relatively high, and effort spent creating the data is not taken into account in this regard.

Patents: As of the date of publication, there is no relevant French case law.

However, there are three cases where Output of generative AI is protected:

1. Output is protected where a user uses AI as a mere tool and the user makes the creative decisions in a way that provides such concrete specifications that the design of the work is already determined. The user is then considered to be the author of the output.
2. An Output may be protected as part of an overall project if the autonomous design by the AI system of a part of the work is used as an artistic element.
3. Some neighboring rights to copyright are designed to protect investments, and are not linked to human intellectual performance. For AI products, protection by these neighboring rights can be considered, in particular, for sound recordings, databases, press products and moving images.

Patents: German patent law is neutral with regard to technical aids used in the process of making an invention. Therefore, computer-generated inventions are patentable under applicable law. The inventor of a computer-generated invention could be the User of the AI system. In addition to the User, the Provider or the producer of the training data may be a co-inventor if he or she has contributed significantly to the generated invention by designing the AI.

However, AI cannot be named as inventor on patent applications.

The Legal Board of Appeal of the European Patent Office has issued its decision in cases J 8/20 and J 9/20 (AI DABUS designated as inventor), which confirmed that, under the European Patent Convention (EPC), an inventor designated in a patent application must be a human. Both the naming as inventor and the transfer of the rights to the invention (and in particular, the right to the patent conferred on the applicant) are considered not possible for an AI system because it lacks legal personality, according to the Legal Board of Appeal.

In its decision for Germany, 11 W (pat) 5/21, the Federal Patent Court also did not permit the sole naming of the AI system, but allowed a co-naming of the AI system (without having to answer the question of the legal capacity of an AI system).

As such, the ownership of an AI-generated work would probably depend on the contribution of a generative AI system’s creators and the User in arriving at the work in question.

Copyright: Under Dutch law, Output solely generated by independently working AI systems, such as ChatGPT, is unlikely to currently benefit from copyright protection.

In order to satisfy the conditions for copyright protection, a work must be the maker’s own intellectual creation, meaning it must have its own, original character and must bear the ‘personal mark’ of the maker. The work must be the result of creative human labor and creative choices of the maker, making it a product of the human spirit.¹⁴

In the case of self-learning AI systems, although such systems have to be fed by the developers of the system with material on the basis of which the learning process can take place, the Output is created in such a way that no maker can be identified as having left their personal mark on it. As such, the conditions for copyright protection are generally not met.

However, if human creative involvement on the Output is extensive enough that creative choices leave a personal mark on the Output, copyright protection will apply. The threshold for such involvement is low, meaning that limited adaptation to the Output (say, in an amended version) would generally allow it to benefit from copyright protection in favor of the person who carries out such adaptation. In addition, if the User post-processes the Output even in a modest matter, that may result in the edited result being protected under copyright if it can be demonstrated that sufficient human creative choices were made.

Case law on determining whether works are protected by copyright is highly fact-specific and we are not aware of any case law in respect of Output from AI models. Although it is therefore difficult to say, it may nevertheless be possible that unaltered Output directly generated by an AI model could also enjoy copyright protection. This may be the case if the human creative involvement through his or her input is sufficiently concrete and detailed in such a way that the Output is deemed to have been created by the AI model under human ‘guidance and supervision.’

Rights in data: No property right or intellectual property right subsists in data per se under Dutch law.

Where the Output contstitutes a database, this can be protected under the database right. The Database Act¹⁵ protects a database in which a substantial investment has been made (either financial or in time and effort) to obtain, control or present the contents of the database. The rightholder is the (natural or legal) person who bears the risk of the investment made for the database – such person is given the exclusive right to grant permission for the retrieval or reuse of a substantial part of the contents of the database (systematic retrieval of small parts of the database is also not allowed).

It remains unclear the extent to which a Provider or User of a generative AI system could benefit from the database rights in practice. The threshold for demonstrating substantial effort is relatively high and the effort spent on creating the data is not taken into account.

Patents: Similar to Germany, Dutch patent law is also neutral concerning technical aids used in the process of making an invention. AI-generated inventions are therefore, in principle, patentable. However, there is no Dutch case law with regard to the question as to who may be considered the (co-)inventor of inventions resulting from Output (for example, the User and/or the Provider).

Copyright: Singapore copyright law distinguishes between authorial works, which must have an author in order to enjoy copyright protection, and non-authorial works, which (as the name suggests) need not have an author. There is no statutory recognition of the subsistence of copyright in a computer-generated authorial work. For an authorial work to be protected by copyright in Singapore, it must be created by a human author (or authors) and be original and not de minimis. Arguably, if AI is treated as a mere tool used by the human author to create the work, such that there is an engagement of the human intellect (for example, an application of intellectual effort, creativity or the exercise of mental labor, skill or judgment), the AI-generated authorial work may be protected under copyright with the human as the author.

On the other hand, if there is a high degree of automation, the requirement of originality may not be met and copyright will not subsist in the work. Accordingly, for Outputs that are wholly generated by the generative AI system without any engagement by an identifiable human author, no copyright would subsist.

As for non-authorial AI-generated works such as sound recordings and films, these may potentially be protected by copyright if a maker can be identified. This could be the company that owns the master sound recordings or made the arrangements for the film. However, the underlying components of the sound recording or film, such as the music, lyrics and scripts, are authorial works. They would not be protected by copyright unless a human author was involved in their creation.

Patents: Under Singapore law, the inventor of the patent has to be its ‘actual devisor’ and this must be a natural person. Similar to copyright, if there is sufficient human intervention vis-à-vis the Output, there could arguably be sufficient human connection that a natural person can be specified as the inventor.

The Output of generative AI might constitute a ‘computer generated work’ in terms of the Copyright Act 1978.

Section 1 of the Copyright Act 1978 defines the ‘author’ in relation to a literary, dramatic, musical or artistic work or computer program which is computer-generated as ‘the person by whom the arrangements necessary for the creation of the work were undertaken.’

Payen Components did not provide clear guidance as to who would constitute the person who ‘made the arrangements necessary for the creation of the work.’ In order to determine ownership, it is first necessary to determine who the author is. Accordingly, there remains doubt as to whether a User could be considered author of the Output if it is a computer-generated work.

Copyright: By statute, copyright can subsist in computer-generated works. The author of a computer-generated work is ‘the person by whom the arrangements necessary for the creation of the work are undertaken.’¹⁶

Although computer-generated works technically have no human author, the AI system will be considered to be a tool of the person making the necessary arrangement for the creation of the work.

For works that are wholly generated by the generative AI system, the author(s) will likely be the generative AI system creators; for works that are created using Prompts input by the User, the User might be deemed to be the author of the work in its own right, or jointly with the generative AI system creators, depending on exactly what level of effort has been expended and by whom. There is as yet no English case law in relation to AI systems in relation to these issues. In the context of video games, in Nova Productions v Mazoomi,¹⁷ the English Court of Appeal held that the creators of a video game were the owners of the copyright in screenshots taken within the game by the game’s players.

Rights in data: No property right or intellectual property right subsists in data per se under English law. If the Output could be said to be a database, it is possible that the contents of the database could be protected by the sui-generis database right, provided that there has been substantial investment in obtaining, verifying or presenting the data. The owner of the database rights is the maker of the database, who is the party that takes the initiative and assumes the risk of obtaining, verifying or presenting the contents of the database. The maker can be a human or a company. The owner of any database rights, as between the User and the creators of the generative AI system, will depend on who has expended the effort in obtaining, verifying or presenting the data.

In practical terms, only rarely will there be the requisite investment in obtaining, verifying or presenting data (one reason for this is, where the data in the database has been created by the AI system, the investment in creating that data is excluded from assessment).

Patents: The inventor of a patentable invention is ‘the actual deviser of the invention’¹⁸ and no provision is made for computer-generated inventions.

The Supreme Court in Thaler v Comptroller General of Patents, Designs and Trade Marks [2023] UKSC 49 held that:

• Inventor: a machine powered by AI could not be an ‘inventor’ for the purposes of the UK Patents Act 1977 (an inventor had to be a natural person).
• Ownership: The Patents Act did not confer a right to obtain a patent for an invention generated autonomously by an AI-powered machine, and still less did it confer such a right on a person who claimed purely on the basis of ownership of such a machine.

This case, though, was on a peculiar set of facts in which the owner (Dr. Thaler) of the AI system named the AI system itself as the inventor in the relevant application. It does not mean that a human inventor cannot successfully apply for a patent in relation to an invention that has been devised with the assistance of an AI system (and indeed in the Court of Appeal, one of the judges suggested that the problem at hand could have been avoided if Dr. Thaler had named himself as the inventor in the application).

Copyright: Copyright can protect only material that is the product of human creativity. Most fundamentally, the term ‘author,’ which is used in both the US Constitution and the US Copyright Act, excludes non-humans. The US Copyright Office’s registration policies and regulations reflect statutory and judicial guidance on this issue.

Generally, individuals who use AI technology in creating a work may claim copyright protection for their own contributions to that work. Works that are solely created by AI technology are not able to be protected by copyright.

In the case of works containing AI-generated material, the Copyright Office will consider whether the work is basically one of human authorship, with the computer or other device merely being an assisting instrument, or whether the traditional elements of authorship in the work (literary, artistic, or musical expression or elements of selection, arrangement and more) were actually conceived and executed, not by a human but by a machine. The answer will depend on the circumstances, particularly how the AI tool operates and how it was used to create the final work.¹⁹

An example of the Copyright Office conducting such an analysis was issued in February 2023 – this one, a 13-page report in the context of a work created using an AI platform known as “Midjourney.”

Patents: The US Constitution gives Congress the power ‘[t]o promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries.’²⁰ Recently, Thaler v. Vidal raised the issue of whether an AI system can be named as an inventor on a patent in the absence of a human counterpart. The United States Court of Appeals for the Federal Circuit (the Federal Circuit) held that it cannot – only human beings are recognized as inventors under US patent law.²¹

Specifically, the Federal Circuit:

• Held that 35 U.S.C. 100(f) defines an inventor as ‘the individual or, if a joint invention, the individuals collectively who invented or discovered the subject matter of the invention.’
• Found that, based on Supreme Court precedent, an ‘individual’ ordinarily means a human unless Congress provided some indication that a different meaning was intended (citing Mohamad v. Palestinian Auth., 566 U.S. 449, 454 (2012)).
• Explicitly declined to address ‘the question of whether inventions made by human beings with the assistance of AI are eligible for patent protection.’

Rights in data: There is no independent statute for protection of data per se. Traditional intellectual property protections may, in certain circumstances, provide some measure of protection.

Data are:

• Considered ‘facts’ under US law.
• Are not copyrightable because they are discovered, not created, as original works.

Although data itself cannot be protected by copyright, copyright may exist in the compilation of the data. The creative arrangement, annotation, or selection of data can be protected by copyright. Utility patent protection may be available if the collection of data is associated with new, not obvious and useful machines, processes, articles of manufacture or improvements of the same. To the extent data is secret and provides the owner with a commercial advantage vis a vis others in the marketplace, data collections may qualify for trade secret protection.