Human, All Too Human: Authorship in Copyright Law

A recent ruling in China held that images created by AI can be copyrighted. This puts China in stark contrast with the USA’s approach, which holds that human authorship is a bedrock requirement of copyright. But what does ‘human authorship’ actually mean? When I take a photo by accident, that photo is covered by copyright law. When I set a camera on a motion sensor in a jungle, and a passing tiger triggers the sensor, then I can claim copyright over the resulting photo. It’s worth considering whether these situations, in which ‘human authorship’ is arguably minimal, but human technology is key, are any different to a situation in which I use a human technology like ChatGPT to create an image after inputting a prompt.

Imagine that the prompt I entered was an incredibly detailed piece of writing, which tasked the A.I. with creating a never-before-seen landscape rich with surrealist scenes beyond most humans’ wildest imaginations. Surely the resulting output is a product of greater human creativity than the photo triggered by a camera on a motion sensor, or the photo I accidentally snapped on holiday? And given that copyright law is intended to protect and encourage human creativity, surely the output of that A.I. ought to be copyrightable?

In the US, the requirement for human authorship means the written words of the prompt in the above scenario could be copyrighted, but the resulting artwork could not. Copyright is vested in the work upon creation, and unlike other IP rights, it doesn’t need to be registered (though the US is unique in requiring registration if you want to litigate your copyright). That means that although the prompt would be protected as soon as you wrote it, anyone could come along and nab the resulting image, use it themselves, and not attribute its creation to you. A pretty gut-wrenching prospect.

If you think that’s unfair, what do you think of this situation? Say you provide a visual artist with detailed instructions on how to draw an image, pulling out all the stops and injecting your own creative energy. The artist then creates the image exactly as instructed, adding none of their own flair. Assuming the artist is a freelancer (and not employed by a corporation that would claim copyright), then the copyright in the work would be vested in the artist, and not in you. This is because copyright law protects expressions, not ideas. You provided the idea. The artist provided the expression.

But hang on, didn’t the generative A.I. provide the expression in the earlier example? Why isn’t that expression protected by copyright? The answer, in the USA at least, is that the expression must be authored by a human (in the UK, this isn’t a requirement). If you think it’s absurd that the output of a camera, but not of an A.I., should be eligible for US copyright protection, then you’re not alone. I would argue that if the written prompt is sufficiently original to warrant copyright protection, then the resulting output should be copyrighted too.

For example, if I type ‘banana’ into a generative A.I., then that prompt is not copyrightable. An expression must be original for it to qualify for copyright. ‘Original’ in this context means that:

a) it is not copied from another work; and

b) that more than a minimal amount of skill, labour and judgement must have been used in creating the work (this is why names and slogans alone generally don’t get copyright).

So whereas the prompt ‘banana’ wouldn’t get copyright, and neither would the subsequent output, I would argue that a long, detailed, well thought-out, and imaginative prompt would be sufficiently original and draw on more than a minimal amount of skill, labour and judgement, in order to attain copyright protection. Therefore it seems clear to me that the resulting output should be protected too, even if it was generated by an A.I. If photography counts as a copyrightable artwork, even if it is strictly brought into existence by technology operated by a human, then so should the output of generative A.I., where the prompt was sufficiently original to warrant copyright.

We’ll see whether the US approach softens to allow for examples like the above. Meanwhile, one upshot of the US’s decision not to grant copyright to any A.I.-generated art is that those with highly creative ideas for images are likely to hire a human artist to create them, in order to ensure that the image is copyrightable and can therefore be profited from.

So protection for human artists comes at the expense of prompt-engineers-come-artists. This may, after all, be a trade-off worth keeping.