Gen-AI output: why the US Copyright Office is right about human authorship

While matters such as commercial confidentiality, data protection, bias and reliability have been raised, much of the debate about what the policy response to gen-AI should be has been led by the creative industries lobby. It has been calling for more to be done to ensure that developers of gen-AI tools compensate them fairly when their content is used to train and operate those systems. For their part, AI developers have pushed back against the notion that what they are doing infringes copyright law. Policymakers globally are grappling with how to balance the competing interests.

However, separate AI and copyright questions have also arisen: to what extent is gen-AI output copyrightable, and does the law need to change to make it easier for gen-AI output to benefit from copyright protection?

These are questions that the US Copyright Office considered as part of a wide-ranging inquiry into the intersection of copyright and AI it gave notice of in 2023. 

In the second of a three-part series of reports published in January, the Copyright Office was explicit in its view: yes, gen-AI output is copyrightable “where AI is used as a tool, and where a human has been able to determine the expressive elements they contain”; and no, copyright law should not be expanded to provide more protection for pure gen-AI output.

The conclusions drawn by the Copyright Office are overall sensible, because they reflect the purpose of copyright law and, broadly, the legal position as it is understood in the EU and UK. Below, we explore what the Copyright Office considered in more detail.

Inputs, or ‘prompts’

In assessing whether gen-AI outputs are copyrightable, the Copyright Office reflected on the fact that US copyright law requires – among other criteria – for there to be human authorship of works for those works to be eligible for copyright protection.

It examined, among other things, whether the output from gen-AI systems could be said to be authored by humans, and therefore potentially qualify for copyright protection, on the mere basis that humans have provided ‘prompts’ to those systems.

According to the Copyright Office, the answer is ‘no’.

It said: “Given current generally available technology, prompts alone do not provide sufficient human control to make users of an AI system the authors of the output. Prompts essentially function as instructions that convey unprotectible ideas. While highly detailed prompts could contain the user’s desired expressive elements, at present they do not control how the AI system processes them in generating the output.”

The Copyright Office went on to say that it is theoretically possible that “AI systems could someday allow users to exert so much control over how their expression is reflected in an output that the system’s contribution would become rote or mechanical” – in essence, that the output could be said to be of human authorship. However, it said the evidence of how the systems operate today indicates that this is not currently the case.

“Prompts do not appear to adequately determine the expressive elements produced, or control how the system translates them into an output,” it said. “The gaps between prompts and resulting outputs demonstrate that the user lacks control over the conversion of their ideas into fixed expression, and the system is largely responsible for determining the expressive elements in the output. In other words, prompts may reflect a user’s mental conception or idea, but they do not control the way that idea is expressed. This is even clearer in the case of generative AI systems that modify or rewrite prompts internally. That process recasts the human contribution – however detailed it may be – into a different form.”

‘Expressive inputs’

The Copyright Office also said the position is different in respect of gen-AI outputs based on “expressive inputs”. That is, where text, images, audio, video, or a combination thereof input to a gen-AI system is “substantially retained as part of the output”.

In cases where the human input is itself copyrightable and that work is “perceptible in the output”, that person will be “the author of at least that portion of the output”, the Copyright Office said, before going on to explain how that part of the output could benefit from copyright protection.

The Copyright Office elaborated: “[In those cases, the human’s] own creative expression will be protected by copyright, with a scope analogous to that in a derivative work. Just as derivative work protection is limited to the material added by the later author, copyright in this type of AI-generated output would cover the perceptible human expression. It may also cover the selection, coordination, and arrangement of the human-authored and AI-generated material, even though it would not extend to the AI-generated elements standing alone.”

Modifications to gen-AI output

The Copyright Office further considered whether changes that a person makes to output generated by AI systems is copyrightable. It said that in some cases it might be, again reflecting on the concept of human authorship.

“Generating content with AI is often an initial or intermediate step, and human authorship may be added in the final product,” it said, noting that some AI platforms offer tools enabling users to “select, edit, and adapt AI-generated content in an iterative fashion”, and how other gen-AI systems have tools enabling users to “exert control over the selection, arrangement, and content of the final output”.

The Copyright Office continued: “Unlike prompts alone, these tools can enable the user to control the selection and placement of individual creative elements. Whether such modifications rise to the minimum standard of originality required [to attain copyright protection] will depend on a case-by-case determination. In those cases where they do, the output should be copyrightable.”

It went on to say that the mere fact that a human-authored work contains AI-generated content does not preclude that work from benefiting from copyright in general, citing the example of a film with AI-generated special effects as an example – even though the AI effects themselves would not be copyrightable in isolation.

Comparisons with other jurisdictions

The Copyright Office further considered whether US copyright law is out-of-step with the law in other jurisdictions in so far as it relates to AI. It reviewed legislation and case law in jurisdictions such as Korea, China, the EU, UK, Australia and Canada, determining that there is not yet a settled legal position globally on the copyrightability of AI output.

It said: “Although some level of consensus on the need for human authorship appears to be emerging, and most countries have so far continued to apply existing law, it is clear that views are still being formed. It remains to be seen how copyrightability standards will be interpreted and applied. The Office is closely monitoring developments abroad and evaluating how other countries’ evolving approaches may ultimately overlap or differ from our own.”

While the Copyright Office’s summary of the legal position in the EU closely reflected the position it articulated in its report in respect of the US legal position, it did highlight a significant difference notably between UK and US copyright law.

UK copyright law, unlike US or EU copyright law, provides scope for computer-generated works to benefit from copyright protection where there is no human author of the work. As we have previously explored, however, the fact the law requires no human author may itself severely restrict the scope for AI-generated output to benefit from copyright protection as a computer-generated work, given the human involvement in prompts.

The law on computer-generated works is currently the subject of UK government review via a broader consultation on copyright and AI that only closed late last month. In that consultation, the government invited views on potentially reforming the law on computer-generated works but noted a lack of evidence about use of the current provisions. The government said its preference would be to remove the protection for computer-generated works from UK copyright law if the consultation reveals “insufficient evidence of [its] positive effects”. It is not yet clear when the government will announce next steps in response to the feedback from its consultation.

No case for change

The Copyright Office was clear in its view that “the case has not been made” for US copyright law to be altered to provide “additional protection for AI-generated material”.

It reflected on the purpose of the copyright clause in the US Constitution, which it characterised as being to incentivise human authors to “promote progress”. It further dismissed calls for separate ‘sui generis’ rights of copyright for gen-AI output to be provided for in law, to incentivise AI development, as unnecessary and queried whether doing so would promote progress.

It said: “We share the concerns expressed about the impact of AI-generated material on human authors and the value that their creative expression provides to society. If a flood of easily and rapidly AI-generated content drowns out human-authored works in the marketplace, additional legal protection would undermine rather than advance the goals of the copyright system. The availability of vastly more works to choose from could actually make it harder to find inspiring or enlightening content. Indeed, AI training itself is reportedly reliant on human-generated content, with synthetic data leading to lower-quality results.”

The Copyright Office concluded that it believes that “existing legal doctrines are adequate and appropriate to resolve questions of copyrightability”.

“Copyright law has long adapted to new technology and can enable case-by-case determinations as to whether AI-generated outputs reflect sufficient human contribution to warrant copyright protection,” it said. “In many circumstances these outputs will be copyrightable in whole or in part – where AI is used as a tool, and where a human has been able to determine the expressive elements they contain. Prompts alone, however, at this stage are unlikely to satisfy those requirements.”

The Copyright Office said it would continue to monitor technological and legal developments to evaluate whether a different approach is needed.

Why I think the US Copyright Office is right

At the heart of copyright is the concept of a genuinely human creation. By definition, AI is not human and so, there is no way copyright law can adapt to recognise purely AI-generated content as eligible for copyright protection. That is our firm conviction. That said, artists and authors have always adapted their creative conduct in consideration of the technical means available.

As an example, the development from cave painting to graffiti is also a technical one. As long as the technical means are of auxiliary nature, there is still enough room for human creativity. If AI-based tools are used to produce some initial drafts with creativity stamped into both the prompting of those tools at the outset and the refinement of the output generated by AI, the overall process may lead to content that can be protected by copyright.

What we find difficult is where the creative part of human intervention or steer stops at the point of prompting information into the system. In such a case, AI is likely to “take over” the most important part of the creative process. That scenario cannot lead to output that involves sufficient human creativity and genuine originality.