Can AI Own an Idea?

This picture ignited controversy when it became one of the first AI-generated images to win a major art prize. But when the artist tried to register it with the U.S. Copyright Office, they denied his request, saying it contained too much AI material. This decision has sparked a debate. Who owns an idea?

An idea is something such as a thought or conception that is the product of mental activity. So who owns that mental activity? Is mental activity just restricted to humans or also machines? If you expect a clear cut answer like a yes or a no, sit tight. The jury is still out.

The concept of protecting ideas has actually changed drastically over time. If you think about the Romans and Greeks, it was more about protecting a collective knowledge, sharing the knowledge. But then the Renaissance happened and the focus was put on the individual, recognizing that individual’s creativity and works. That actually lay the grounds for what is today the intellectual property system that we live by.

The first IP laws were enacted because the Industrial Revolution was based on inventions, on ideas that people wanted to protect. Today, our laws, the copyright, the trademark, the patent laws are there to protect creativity, originality. But in an age of artificial intelligence, are they doing that or not?

The final adoption of the Artificial Intelligence Act today is a historic moment. This landmark legislation is a significant step forward in our commitment to creating an ethical future for artificial intelligence in Europe.

Laws right now are very different. You have the US and you have Europe. Europe right now is actually the most restrictive system since August of 2024, when the EU act was passed. And it is based on different risk levels from minimal to prohibited risk. Actually what governments and legislators, law firms, institutions are working on worldwide right now is twofold.

On the one hand, is protecting the owners of copyrighted materials from infringement on their rights. And on the other side, how to actually protect AI-generated content. Regarding the IP rights, the prompt writers – the users – are actually the owners of the prompt if there is sufficient originality, if the human behind the prompts has really given it a lot of thought.

However, they are not the owners of the output. And I will give an example. It’s called Zarya of the Dawn, and the author of this graphic novel actually wrote the novel, and the Copyright Office said that she could have the rights to the text and to the layout of the pictures, but not to the pictures, because they had been produced by Midjourney.

The output of something generated just by the machine is actually not copyrightable, because it has to be linked to the principle of originality, which means that it has to be linked to a human being. A very interesting subject is whether there is protection for using, for training purposes, copyrighted materials. For instance, I may have written a book that you may have access because you have paid for that.

Then you put it out there., you feed it into the machine, it’s gone forever. I’m not going to benefit from that. Well, it’s not even clear at this stage if that is actually legal or not. And for instance, now there are major lawsuits open in the States. One is The New York Times that has actually sued OpenAI, the makers of ChatGPT and Microsoft, because they back OpenAI, because they said that there is a clear IP infringement, because ChatGPT has been fed a lot of information coming from The New York Times without its express consent.

The other one, in visual terms, is, Getty Images because it claims that Stable Diffusion has actually taken 12 million pictures from its site without giving it either credit or financial compensation. So now what they’re doing is that you have what’s called opt-out clauses that you put in your website that actually protect you. If you are able to discern that somebody has been using your information.

How do you know that they’re not using it? It’s practically impossible. On the other hand, the providers of the AI systems are also being very careful. And you look at ChatGPT’s terms and conditions, you will find a lot of talk about transparency, fair use, copyrighted materials. Some very large institutions, what they’re doing now to prevent this copyright infringement issue from their employees is actually creating their own LLMs, their own large language models.

And so they feed into those models what they deem necessary for the employees. It works a little bit like an intranet, if you wish. Only they have access to that, not external sources. I think people are starting to embrace the technology, to not fear it. For instance, they are creating new ways of licensing so that you can benefit from other people using your copyrighted materials.

They are now, thinking about how to protect actual output by machines because at the end of the day, ChatGPT is only a tool. Speaking about the ownership of ideas, it may come a day when the ideas generated by ChatGPT will be its own.