Who Owns the Copyright on AI Generated Content, like ChatGPT?

There’s a possible risk of infringement

Many Medium writers have written about ChatGPT and their experiences feeding it questions and prompts. In fact, it looks like a really exciting tool, with the potential to put writers out of a job — although we’d still need an army of fact checkers, for anyone who can be bothered with little details like accuracy.

Some people describe it as, “the most important development since the invention of the printing press or the splitting of the atom,” says Joe McKendrick on Forbes. Whether it’s one of the ‘most important developments of our time’ remains to be seen, but it’s certainly worthy of our attention.

Let’s start with some of the problems. AI software can produce great content, or mediocre content. It depends on how good it is and how well it’s trained. ChatGPT is certainly among the better AI programs.

However, there is a risk that sometimes, AI can be short on — or completely devoid of — fact checking and accuracy. Essentially, some of what it produces may sound good and be confidently relayed, but it’s just plain wrong. So as a user, you need to check what it says, because it’s not necessarily accurate.

For those people who want to play around with questions and answers, there are no issues around copyright, but if you want to use the copy it produces in an article, the question of who owns the copyright isn’t clear cut.

Bern Elliot, VP Analyst at Gartner, said about ChatGPT: “The model is trained on a corpus of created works and it is still unclear what the legal precedent may be for reuse of this content, if it was derived from the intellectual property of others.” There’s also, notably, no citation or source provided by the bot.

Can works created by non-humans be copyrighted?

According to Forbes, work created by a non-human cannot be copyrighted…

For a work to enjoy copyright protection under current U.S. law, “the work must be the result of original and creative authorship by a human author,” says Margaret Esquenet, partner with Finnegan, Henderson, Farabow, Garrett & Dunner, LLP. “Absent human creative input, a work is not entitled to copyright protection. As a result, the U.S. Copyright Office will not register a work that was created by an autonomous artificial intelligence tool.”

So it would seem that the answer is clear cut — AI content cannot be copyrighted. But actually, the answer is rather murkier than it might seem at first glance.

A court case about whether a monkey could own the copyright to a photograph caused a massive legal battle between 2016 and 2018, leaving the photographer claiming he’d lost £10k in revenues from sales and was unable to pay his attorney, as the case went through court cases and appeals.

While the court eventually settled on the side of the photographer, because he’d sent up the shoot, so was significantly instrumental in the picture’s creation, it was a very messy and widely disputed case about whether a non-human can hold copyright. The end result, at this time, being ‘no’. But that didn’t stop the case from going to court, dragging on for three years, and almost bankrupting the photographer.

At present, the rules are as follows:

US law says that AI works are either ‘public domain’ upon creation, or a ‘derivative work’, created from the materials the AI was exposed to while it was being trained. In the case of ‘derivative works’ there could be an issue with copyright infringement if the end piece is too similar to the original works upon which the software was trained. And there in lies the problem.

AI programs do not cite sources, so you cannot check to see whether the finished work is too similar too the original work that inspired it. The only way to check might be to run some lines through a plagiarism checker, or through Google, but that’s certainly not fool-proof, nor is it guaranteed to bring up infringements.

A Forbes journalist speaks in some depth to lawyers about this here.

What if two users get the same response?

Sometimes the same response may be duplicated by AI software for two different users. In this instance, users must understand that neither of them can claim copyright of the content. All users should understand that the output of the software may not be unique, and may have also been given to other users.

If the software creates an exact copy of copyrighted material, there’s also a murky question of who is held responsible for copyright infringement — the owners of the software (such as ChatGPT), or the users of the software, such as you and me. It’s the kind of question that is likely to be thrashed out in court in years to come.

The bottom line

The bottom line is that if the work created by the bot is unique, it should be in the public domain — so no one owes the copyright. However, it’s virtually impossible for a user to ascertain, with any degree of certainty, if the work created is indeed unique, or whether it infringes on someone’s copyright from the original training source.

In that case, there remains a risk of infringement, which will probably play out in courts at some time in the future as people and companies adopt these assistive technologies more, in efforts to keep their costs down.

How to reduce the risk of legal action

If, as writers, we use AI as a tool for inspiration only, and then rewrite or heavily edit our final copy, it reduces the risk of copyrighted work being inadvertently used in our own creations.

If you rewrite the work, you will not use the original text supplied by the bot, and you can still apply your own style and your own thoughts to the finished piece of work. It’d be a lot like taking inspiration from another article, and then writing your own piece on the subject.

Taking this approach will probably make for a better piece of work and, of course, you can do that all essential fact-checking and cite sources, as you go.

Doesn’t that erode the benefits of using AI, though? Yes, it might, to some extent. But it’s a way to reduce the risk, while putting your own stamp on the end product to give it extra value.

Ultimately, whether you’re happy to use AI, and accept the risks of copyright infringement associated with it, will depend upon your appetite for risk. How much of a problem this ends up being in the future remains to be seen.

These programs might do a good job of rewriting content, so copyright infringement isn’t necessarily inevitable. Although you’d think some plagiarism would creep in — in the sense that the bot is regurgitating someone else’s ideas without crediting the source.

© Susie Kearley 2023. All Rights Reserved.

More of my stories on copyright can be found here…