Who owns AI-created content? Right now, 'nobody knows,' expert says

By Seda Sevencan

ISTANBUL (AA) – As artificial intelligence (AI) advances, its intersection with copyright law remains fraught with legal uncertainty, raising questions about fair compensation for creators and the boundaries of intellectual property.

At the 2nd Symposium on News Copyright and Artificial Intelligence in Media, held at Bogazici University and organized by Anadolu, Guido Westkamp, an academic at Queen Mary University of London and expert in intellectual property and comparative law, underscored the legal ambiguities surrounding AI-driven content.

With AI systems trained on vast datasets, also including copyrighted materials, concerns persist over whether creators will receive fair compensation. Yet, as Westkamp pointed out, the legal framework remains unclear.

"It's a very topical question, and the answer is that at present, nobody knows," he said.

Westkamp explained that current copyright law includes exceptions allowing data mining — rules originally designed for scientific research rather than AI training.

"We have an exception in law for copyright that allows data mining. That wasn't made for AI. It was made for other purposes, and there are two exceptions."

"One for scientific research — so everything that is done for scientific research is permitted. There is another one that is for general purposes, where authors can actually opt out, by basically declaring that they do not wish their works to be trained. That causes a number of additional issues."

A ruling from a lower court in Germany has further complicated the debate. "There is now a very recent decision, one of the first decisions on whether this data mining exception can apply to AI. It's a decision from Germany,” Westkamp explained. “The court said, yes, it can."

"In other words, the data mining exception applies generally, even if the data mining is done to appropriate data, including works, in this case photographs, for the purpose of training generative AI models. If that is the case, then no further issues will arise. In other words, there's not going to be any kind of remuneration."

"It is just free as a matter of freedom of research, i.e. scientific research. Whether that decision will stand is an entirely different matter. It will go up, of course, through the court system."

Whether this ruling will hold up in higher courts remains uncertain. "It will probably end up with the Court of Justice for the EU, where the court will have to address questions of balance," Westkamp added.

He believes that creators should have a right to remuneration and the ability to opt out. "My own personal view is that, necessarily, there needs to be maybe a right to remuneration, predominantly a right to remuneration that ensures authors, journalists, and so on are paid. And at the same time, authors should have a right actually to object," he said. "In other words, to opt out of a system like that."

- Should AI-generated content be copyrighted?

One of the questions in copyright law today is the ownership of AI-generated content. Should it belong to the creator of the AI, the organization using it, or someone else?

"That very much depends, because if it is a fully automated result of some AI process, in my view, no one," Westkamp stated. "Because it is made by a machine, and of course copyright law is based on human creativity, and I think that is very clear.”

“There may be ideas to say it's a kind of investment, but I find that rather far-reaching. When it comes to instances where maybe an artist, or a journalist also, simply uses AI as a tool, then it can be a creative process.”

He dismissed the argument that AI-generated content could be classified as an investment, saying: "There may be ideas to say it's a kind of investment, but I find that rather far-reaching."

However, he made a distinction for cases where AI serves as a tool rather than as the sole creator. "If an artist or journalist simply uses AI as a tool, then it can be a creative process. People have always used technology to be creative. In that case, we still have human authorship," he explained.

"When it comes to instances where maybe an artist, or a journalist also, simply uses AI as a tool, then it can be a creative process. And then there is no distinction between, you know, people have always used technology to be creative in that sense, so if it's used in that sense, then nothing will change. We will still have human authorship," he said.

"But it's clear, I think, if it's fully automated, no authorship. If you're just using it as a way to express yourself, as a tool, if you have a creative idea behind it that you want to express, in that case, I think it's just a bog-standard copyright matter."

For media organizations adopting AI-driven tools, compliance with copyright laws is an evolving concern.

"At the moment, this is just really a matter of transparency in the EU," Westkamp explained.

"So the AI Act is basically saying you are fine using AI, but you would need to follow certain steps in terms of being clear on the works you've used. So apparently that means having records and things like that. Whether there will be more in future is a different matter."

- Broader implications of AI and copyright law

Beyond the immediate legal challenges, Westkamp sees AI as a fundamental test for intellectual property law in the digital age.

"In general, maybe the discussion is very compartmentalized. In copyright law in particular, or even in patent law, can this be creative, can this be inventive? Yes, you can discuss all that, but there is a much broader concern, which I think has to do with the relationship between humanity and robots, so to speak," he said.

He raised philosophical questions that may shape future legal debates. "And the question of, you know, what does that have to do with fundamental rights? Can, for instance, AI be a bearer of fundamental rights, a bearer of freedom of opinion and freedom of art and all these rights that have to do with communication? And that very much is, in my view, the more fundamental question that needs to be addressed in the years to come, so that there is clearly no immediate answer to that."

"We will see in 20 years how things have turned out. I think it is very premature," said Westkamp.