GenAI is Not a Legal No Man’s Land

While many claim that GenAI is a legal no man’s land, Aurore Sauviat, a lawyer dedicated to working with creatives and advertising agencies, brings clarity to the issues of IP rights and GenAI use. While there are many questions about IP and the use of GenAI, Aurore Sauviat, a partner at Lawderis Avocats, explained to Visionary Marketing, there is no such thing as a legal loophole and answers are out there. Does creating with AI necessarily lead to forgery? Do we have the right to commercially exploit GenAI results? Can I claim copyright over what I have created using generative artificial intelligence tools, not to mention questions of image rights for individuals, living or dead? And finally, who is responsible if an AI chatbot makes mistakes? Aurore addressed the topic of copyright, image rights and liability issues in the use of AI, providing clear answers. Reassuring professionals who shudder at the thought of pressing the return key after entering their prompts. Aurore’s answer: GenAI is definitely not the legal no man’s land that laymen think it is. 

IP Rights and GenAI: There is No Legal No Man’s Land

IP rights, copyright and GenAI

Aurore Sauviat warned us: “One reads a lot of misconceptions in the media and on social platforms. There is such a lack of understanding of copyright issues that people are mixing up subjects that have nothing to do with one another.” On the one hand, it’s about the liberties taken by publishers of AI solutions to train their algorithms, and on the other hand, how these tools are used to produce content and images in particular.

AI training by publishers

“There are already 50 lawsuits in progress in the United States on this subject and a couple in the UK. These disputes are beginning to surface around Europe, with a few in Germany and a new one in France.”

While most users and authors might think that the scraping of content and works of art by GenAI platforms equates to pure and simple intellectual property theft, the underlying question of IP rights is far less straightforward, the lawyer explained to Visionary Marketing.

Artificial intelligence platforms are indeed resorting to “certain exceptions in copyright law that allow you not to ask permission from the creator in specific cases.” This is known as “fair use” in the USA.

Definition of fair use in the USA

Fair use permits a party to use a copyrighted work without the copyright owner’s permission for purposes such as criticism, comment, news reporting, teaching, scholarship, or research.

Copyright Alliance, “What is Fair Use“

Discussions of this topic have led to multiple emerging lawsuits, but, “they are extremely fragmented because they relate to very specific disputes,” explains Aurore. Be that as it may, Aurore states, “the position adopted by AI publishers is not to pay for what they have ‘scraped’ and to use another exception in Europe. This is the “text and data mining” (TDM) exception, which was introduced by Directive (EU) 2019/790 on copyright in the digital single marketMarket definition in B2B and B2C - The very notion of "market" is at the heart of any marketing approach. A market can be defined... (DSM). 

This exception was “unfortunately created in a 2019 copyright directive,” explains Aurore.

At that time, LLMs had not yet been widely adopted. Negotiations initiated by the major content producers, the “majors” SACEM and its German counterpart (GEMA) did not result in an agreement on remuneration for creators whose works had been used without authorization. The reason why these negotiations were unsuccessful, explains the legal expert, is that the “aim of AI publishers is not to pay or to pay as late as possible.”

Does using a tool that infringes copyright make you a thief?

Just because you have used a tool that has infringed copyright does not mean that you have “automatically infringed copyright,” explains Aurore. “Because the notion of infringement is not assessed in a general way,” and this explains why authors find it very difficult to assert their rights.

As long as you use an artificial intelligence tool to create, using a genuine creative process and produce a result that does not resemble existing physical works of art that can be named, denominated, or specifically identified, then you are not infringing copyright.

You can imagine how difficult it is for artists and content creators to assert rights that are ultimately quite fragile and which, moreover, are not protected at the same speed as that which governs technological innovation. Incidentally, this also applies to us, professional content creators, who have been feeding the Internet for 30 years.

Copyright and commercial use of “works” produced with AI?

There’s the upstream copyright issue, but there’s also the downstream issue: “When I integrate artificial intelligence tools into my workflows, can I, firstly, make commercial use of them, and secondly, can I claim copyright?”

These are still “issues that are wrongly mixed up,” insists Aurore.

General conditions: read the small print

You can potentially claim commercial exploitation without being protected by copyright. So the answer to the above question is an enigmatic “it depends.”

It depends on the terms and conditions of use for the tools that you used. Many of these tools have been made available in freemium mode. Aurore therefore urges us to “read the general terms and conditions of use carefully, because free use does not freely, if at all, allow the results produced to be used commercially.” In some cases, a watermark should be affixed. This was the method chosen by Adobe at the time of the launch of Adobe Firefly in 2023. But as far as I know, these watermarks have since been abandoned.

Law and AI: Claiming copyright for AI?

Second question, separate from the first, “Can I claim copyright on what I’ve created using an artificial intelligence tool?” Here again, the answer is “It depends on the originality of your creation,” explains Aurore.

“Legally, this means that your creation bears witness to the imprint of an author’s personality.” The case law legal language can be explained like so. It means that you are able to demonstrate all of your creative choices. Aurore, who works exclusively with creatives, explains that “when creatives use AI to generate something, it’s not with a simple sentence, a prompt, a quick result exploited directly. On the contrary, they work upstream, making sketches, creating moodboards, reflecting on their work and proceeding iteratively. Finally, they also work on post-production, correcting the raw results produced by the AI. What’s more, they don’t give any brand elements to the AI; they integrate them in post-production. “It is this whole creative process that demonstrates copyright and qualifies the imprint of a creator’s personality.”

Below are some examples of brands integrating GenAI into their ad campaigns. Though the public reaction is less than enthusiastic, the companies seem to stand by their decision.

Despite the major differences between copyright and authors’ rights, the situation between France and the United States is not very different, explains Aurore.

The only difference with the United States is that on the other side of the Atlantic, copyright is registered. When you create a work, you submit it to the US Copyright Office and they tell you straight away whether it’s ‘copyrightable’ or not.

Here, the situation is reversed. Legally, in France, you hold copyright as soon as you materialize your work. And it’s only when there’s a legal dispute over plagiarism that you’re going to have to prove that, on the one hand, you own the copyright and, on the other hand, that the work is original and not plagiarized.

To conclude, if you are a creative, it is your responsibility to “document your creative process and keep evidence of that process. Proof of the creative process can then be used to claim copyright on what you have created using artificial intelligence.”

Always document the creative process, this evidence can be used to claim copyright.

AI and image rights

Similar to claiming copyright, is the topic of AI and image rights. What happens when I decide to exploit the image rights of natural persons, when using AI?

This can occur in voice cloning, the re-use of someone’s image, or personality attributes, whether they are alive or dead. The question from Aurore’s clients is “do I have the right to do this?” For once, the answer is much simpler, the legal expert tells us. “It’s not a grey area at all.”

“As long as you are alive and your permission has not been sought, you can object to the use of your image, your voice, your attitude, the way you walk, your name and surname.”

This is what happened with Scarlett Johansson and the voice of OpenAI. Sam Altman, a big fan of the film Her, asked Johansson to use her voice for the voice of OpenAI. She refused. Altman therefore recreated her voice using AI, however, as the actress was still alive, “she objected to the immensely similar replication of her voice.” 

Under the pressure of the bad buzz, OpenAI backed down. But legally, they could have had an argument to use the generated voice because it was in fact the synthesized voice of another human being, even though it replicated Scarlett Johansson’s.

Open-bar for the deceased!

For deceased celebrities, it’s a completely different story. “As soon as you die, you no longer have any personality rights and your heirs don’t get those rights back either.”

If a celebrity dies, anyone can reuse their image for advertising and promotion. That’s why over the years we’ve seen Gandhi selling Microsoft, cars, or Einstein and Marilyn Monroe selling us pizzas. Even if the usage of artificial intelligence in this way offends the public, it’s not breaking the law.

The case of deepfakes

Then there’s the subject of deepfakes. Everyone thinks it’s forbidden. But it’s not as simple as that.

“Deepfake itself is not banned at all, as it is subject to the principle of freedom of expression.”

A deepfake can therefore be used to create a satirical political cartoon. This type of usage has been going on for a long time, the legal expert confirms. Canteloup fans know it.

“The first constraint is that you have to mention that you are doing deepfake,” explains Aurore, and that “you have reused an AI tool to regenerate someone’s image or voice.” The second constraint is “not to undermine public order.” In fact, over 90% of deepfakes are generated in cases of “revenge porn” or for pornographic purposes.

Beyond these limits, deepfakes are allowed.

AI and liability

Having plenty of experience with companies in the events sector, Aurore covers AI and liability in the final use case. This is particularly important if you are implementing chatbots as part of your business.

Am I responsible for what AI says?

Put simply, the answer is yes, explains the lawyer.

The responsibility can not fall on AI, because it’s neither a natural nor a legal person. Aurore explains, “they’re just providing you with a tool, and you’re responsible for how you use it.” Just as you wouldn’t sue Nikon if you took an illegal photo with a Nikon camera, you can’t sue the publisher of your AI chatbot either.

In conclusion, AI is not a No Man’s Land!

To conclude, Aurore reassured us: “AI is by no means the Wild West.” Legal experts have plenty of answers for you.

“There is a significant body of texts in copyright law, consumer law, on the Influencer Act and the famous AI Act, which allows us to provide answers.” Moving forward, Aurore urges us to “adopt AI tools for [our] businesses, if [we] [want to].”

But this conversation also warns us about the need for legal certainty in order to enable these technologies and implement them in a way that is healthy for our businesses.

AI and the law: it’s not a legal Wild West – Aurore’s presentation at Com’ en Or Day

We also invite you to watch our presentation at this event on March 20, 2025