The copyright status of AI-generated works

As AI platforms like Midjourney, or Dall-E2 are widely adopted the question, that is increasingly being asked is about the copyright position.

It’s possible to create striking artistic images, music, poems and the like using Artificial Intelligence (AI) platforms so people naturally wonder who owns the rights in the output? What are the risks of using the output commercially?

One common trait these AI platforms share is that they first need to learn how to simulate human thinking. So, they feed the machine learning software with hundreds of thousands of images so it can learn, for example, what a cat looks like. These images are scraped from the internet. The AI platform then puts some rules in place by which to decide how to generate an image of a cat when a user prompts the word ‘cat’. Drawing on the hundreds and thousands of images it has been fed as part of its learning, the AI platform generates an image of a cat.

Here’s a video explaining how Dall-e2 works which I gather is different to what you would use Midjourney for.  At the time of writing, I haven’t personally used any of these platforms to test how they work, and how they differ. Presumably, different results are generated each time even if the search query is the same? If each output is unique and another user gets a different output even if their query is identical to yours, it means someone else could only use the same visual images as you by copying yours. Would they be infringing your copyright though?

Is the output infringing copyright?

From a copyright law perspective, the initial use of copyright works by the platform for machine learning is infringing unless the platform used licensed or out-of-copyright works or could rely on a copyright defence, such as use for research purposes. As Midjourney and Dall-E2 are both commercial platforms that take images found online it is theoretically possible that one of their image outputs is recognisably the same as a pre-existing image. If so, that image would infringe copyright in the original. So, if you’re intending to use an unusual, highly distinctive image generated by AI, you may want to do some due diligence to establish that there is no obvious infringement. You would be liable if you were making use of the infringing output. But assuming that is just a theoretical risk and that in practice what is created by these platforms is a mash up of many different images, the next question is whether there is copyright in the output, and if so, who owns the rights.

Computer generated works

Since the 1970s computers have been producing crude works of art. Most of these computer-generated works are protected by copyright in the UK. Unlike other countries, the UK gives copyright to computer generated works which do not have a human creator. Section 178 Copyright Designs and Patents Act 1988 (CDPA) introduced this law in 1980s. The CDPA designates the author of such a work as “the person by whom the arrangements necessary for the creation of the work are undertaken” (s9(3) CDPA). Protection lasts for 50 years from the date the work is made (s12(7) CDPA). The person making the arrangements would generally be the platform creators but could sometimes be the programmers operating the tool.

When the law was introduced, it was designed to do more than protect works created using a computer as a “clever pencil”. Instead, it was meant to protect material such as weather maps, output from expert systems, and works generated by AI. These relied heavily on the creative input of the programmer; the machine was at most an instrument or a tool very much like a brush or canvas. But new tools like Midjourney enable a user to generate works with minimal user input. Does the fact that it is easy to generate an output with minimal creative effort detract from its ownability as a copyright work?

Copyright position

Machine learning algorithms make independent decisions throughout the process. While users set the parameters, the work is generated by the computer program itself – referred to as a neural network – in a process akin to the thought processes of humans. The algorithm determines what the new work looks like. So, in the latest types of AI, the computer program is not just a tool: it actually makes many of the decisions involved in the creative process without human intervention. On that basis, applying UK law, it means that if anyone owns copyright in the output it would be the platform owner rather than the user.   

Literary, dramatic, musical, and artistic works are only protected by copyright if they are “original”. In 1988 when these laws were introduced, “original” meant a work must be the product of the “skill, labour or judgement” of its author. But since then, the concept of originality has evolved. Consequently, there is some uncertainty about how the computer-generated works provision applies. The current approach of the law is that a work must be “the author’s own intellectual creation”. This means it must result from the author’s free and creative choices and exhibit their “personal touch”. It is unclear how these concepts apply to some of the AI generated works. So, it’s argued that a separate definition of originality may be needed.

By designating a human as the author of a work generated by an AI, the UK approach separates authorship and creativity. The creator of the original work is the AI, but the “author” under the law is a person who has not made any creative input to it. This sits uneasily with the modern approach to originality in wider copyright law, where creativity and authorship go hand-in-hand. As computer-generated works have “no human author”, it appears that the concept of “joint authorship” does not apply to works co-created by humans and AI systems.

To the extent that a work is made with assistance from AI but involving human creativity it will be protected like any other work. Copyright protection will protect the output if it’s the human creator’s “own intellectual creation”, and the first owner of the work will be that creator. The AI in these cases may be considered to simply act as a tool which allows an artist to express their creativity. In other cases, although a human may have been involved in the process of generating an AI work, their input will not be considered sufficient to attract protection as a standard copyright work. In those cases, the output may be deemed free of copyright because they are not created by a human author. As such, they could be freely used and reused by anyone.  

Other countries

Creative works qualify for copyright protection if they are original, with most definitions of originality requiring a human author. Most jurisdictions, including Spain and Germany, state that only works created by a human can be protected by copyright.  The laws of many countries are not amenable to non-human copyright.

In the United States, for example earlier in 2022, the Copyright Office declared that it will only “register an original work of authorship, provided that the work was created by a human being.”

The degree of human “intervention” needed for a work to become eligible for copyright protection needs to involve more than merely selecting a single work from a number of works created by an AI. If the platform in question is located in the USA then under UK legal principles the owner of copyright in AI generated works would be the platform rather than users if their actions short of the requisite originality.   

In Europe the Court of Justice of the European Union (CJEU) has declared on various occasions, particularly in its landmark Infopaq decision (C-5/08 Infopaq International A/S v Danske Dagbaldes Forening), that copyright only applies to original works, and that originality must reflect the “author’s own intellectual creation.” This is usually understood as meaning that an original work must reflect the author’s personality, which clearly means that a human author is necessary for a copyright work to exist.

The future

Although there is ambiguity about the status of AI-assisted works the UK Government decided against changing the law on computer generated works after it called for views on artificial intelligence copyright and related rights.

The predominant view was that more time is needed to properly evaluate the options. In the meantime, the law will be kept under review and the Government could amend, replace or remove protection in future if the evidence supports it.

So there is a degree of uncertainty for users wanting to rely on the output of these AI generated platforms for their commercial services. They need to decide what impact the ambiguity around copyright means for them. Bear in mind that the law will take time to evolve. Also, it will take even longer for a coherent international approach between countries to emerge.

Some platforms like Dalle 2 state that the user is the person who owns the copyright. On this view you could argue that the fact that it may be easy to generate an output should not detract from the fact that it’s the user’s juxtaposition of concepts that generates the output. So, it’s appropriate that you own the rights in the resulting output. Decide how you will approach the ambiguity before incorporating AI generated works into your work processes.

Shireen Smith is an intellectual property lawyer whose company Azrights International focuses on brand protection, trade marks and copyright. In 2020 she set up the Brand Tuned podcast, and wrote her third book, Brand Tuned, the new rules of branding, strategy and intellectual property. Having trained in marketing and refined her ideas about brand creation she recently created the Brand Tuned Accreditation Program to teach designers, marketers, lawyers, business advisers and founders how to create distinctive brands using IP strategically. Email Twitter @ShireenSmith or @Azrights or @Brand_tuned.

Image: AI-generated sunset by Felixbw, via Wikimedia Commons, CC BY-SA 4.0.