Simpsons | Robo-Rights: Should AI-generated works be protected by copyright?

Quick Hit

Artificial Intelligence (AI) is impacting the world we live in in a number of ways - and the creative industries have not been spared from this rise of the machines. A much-publicised development is the ability for AI algorithms to create art, poetry, music and even novels, with minimal need for human creative input. This raises many ethical, economic and legal issues – amongst them, whether AI-generated art and other AI works should receive the legal protections extended to human-made art.

The arguments for and against copyright protection of AI-generated art

Under Australian copyright law (and, indeed, most jurisdictions around the world) human effort is required to create a literary, dramatic, musical or artistic “work”. So, paintings by elephants and monkey selfies have famously not received copyright protection. However, some countries are now reviewing their copyright regimes in light of the rise of “AI-art”.

Section 32(4) of the Copyright Act 1968 (Cth) requires that the author of an original work is a “qualified person”, meaning an Australian citizen or a person resident in Australia. Further, in the High Court case IceTV Pty Ltd v Nine Network Australia Pty Ltd (2009) 239 CLR 458, the Court emphasised how, for a “work” to be protected by copyright, it must be the product of independent intellectual effort on the part of the author.

An AI system, however, is not itself a “qualified person”, but rather a tool that is programmed to produce an output based on certain inputs. Other “tools” such as a paintbrush, a musical instrument or software such as Photoshop are all commonly used to create copyright works, but the crucial distinction is those tools are substantially controlled by the person operating them. AI-art may be capable of creating new works, but those works will not be the direct result of a human user controlling the result. In legal terms, there may not be any or sufficient independent intellectual efforton the part of a “qualified person” for the resulting “work” to qualify for copyright protection.

There’s also a moral argument against copyright protection for AI-art: the current copyright system values human endeavour; it incentivises the advance of originality and creativity in our society. The speed and ease at which new AI systems such as ChatGPT and Stable Diffusion can create content may arguably risk undermining investment in original work by human authors and artists.

From an economic perspective, the long term of copyright protection for computer-generated works seems excessive as, compared to human creators, there’s no (or little) cost to the AI-system. Should an AI-generated novel or song, created within a few moments, receive comparable protection to the work of an author or composer who has spent years honing their skills?

However, proponents of AI technology claim that AI-art is the product of a creative process and should be recognised as such through copyright protection. There may be an “art” in how best to instruct an AI algorithm to prepare a new work.

Of course, that presumes that it is easy to identify who the owner of the relevant rights should be: the person who sets the parameters for the output or the programmers and engineers who developed the system or the platform owner through which the art is generated?

That said, one assumes that – as currently appears to be the case - platforms will allocate “ownership” based on end user terms and conditions rather than by reference to how copyright legislation defines authorship and ownership.

However, as it currently stands, if one had commissioned or created a particularly valuable piece of AI-art, the uncertainties around human authorship and ownership may also make it difficult to protect that work from third parties using it without permission.

A third way? AI-specific Copyright Protection

An alternative option is that AI-art could be protected by copyright-like legal framework, specifically designed to address the particularities of AI-generated creations and with a period of copyright perhaps measured in a span of years rather than life-times (akin to design and patent protection periods).

Indeed, the UK copyright law protects “computer-generated works”, granting copyright protection to works generated by computer in circumstances such that there is no human author of the work. The author of a “computer-generated work” is taken to be the person by whom the arrangements necessary for the creation of the work are undertaken.[1] This definition appears problematic, as in many cases it may be unclear whether the users of AI programs who feed in inputs, or the creator and/or owner of the program itself would own copyright in a given computer generated work.

Infringement and liability

The rise of AI-art will inevitably lead to questions of infringement and liability.

AI-art could contain offensive or illegal images. For example, an AI-generated “Seinfeld” show has recently come under fire and was banned from the streaming platform “Twitch” for airing transphobic and homophobic statements. Will the programmer or platform be liable for the harm caused? Or the user of the AI-system whose input request generated the material?

If a piece of AI-art reproduces the whole or a substantial part of someone else’s original material, it will infringe copyright. Also, if the creator of the AI-system copies artworks or books to feed into the AI-system without the permission of the owners of those works, that copying could amount to infringement, regardless of whether the end-product infringes. There are several lawsuits unfolding around the world regarding the input and use of existing copyright material, scraped from the internet to “train” AI systems to create content.

In Andersen v. Stability AI et al., No. 23 Civ. 201 three visual artists are alleging that Stability AI unlawfully used their copyright material without permission to generate new and infringing derivative works mimicking the artists’ unique styles. Stock image library Getty images are also pursuing Stability AI in the UK and the US for allegedly scraping and infringing millions of images without a licence.

It may, however, be much harder to prove or even assess whether a particular piece of AI-art “output” infringes a particular input source – including to assess or prove any “causal connection” between the output work and the input.

This would also mean that downstream users may have no assurance as to whether they have bought themselves a lawsuit if they use a work created through AI rather than more directly or demonstrably created by a human.

This is a very fast-moving area of technology and law. If you are facing any legal or commercial challenges relating to AI-generated content, please don’t hesitate to reach out to Adam Simpson, Ian McDonald, Clare Young or Daniel Anstey .