NFTVF 2024 theme year “Talent in the age of AI” ends in a panel with Nordic experts discussing the legal side of AI

To conclude Nordisk Film & TV Fond's 2024 theme year, "Talent in the Age of AI”, the Fund brought together legal experts from all Nordic countries specialised in copyright and AI for an online discussion on the most pressing issues surrounding AI and rights in the Nordic and global film industry.

The discussion, moderated by Norwegian producer Stein B. Kvae from Paradox, gathered a distinguished panel: Victor Hedvall, lawyer at the Writers Guild of Sweden; Petteri Günther, IP lawyer at DLA Piper Finland; Sigurður Örn Hilmarsson, lawyer at the Icelandic law firm Réttur; Thomas Heldrup, Head of Content Protection and Enforcement at the Danish Rights Alliance; Anna Stabell, lawyer at the Norwegian Film Workers Association; and Lisa Digernes, copyright and entertainment lawyer at Norwegian law firm Bull & Co.

Are our current legal initiatives to restrict AI training on creative work truly effective?

Today, when AI trains on synthetic data or metadata, implementing copyright protections has become increasingly difficult. This raises questions about the legal measures available for copyright holders to prevent their work from being used in AI models – a process commonly referred to as opting out.

“In Iceland, as I assume in many other jurisdictions to date, this data mining training (DTM) has been done without permission, we don't have an opt out mechanism by law in Iceland at the moment, just conventional law provisions of copyright law,” Sigurður Örn Hilmarsson says.

According to Danish Thomas Heldrup, legal progress is being halted due to lack of transparency from the AI companies.

“Without proper transparency into what the AI companies have actually scraped from the Internet, we can't really do anything. And right now, they're not being transparent.”

Swedish Victor Hedvall is on the same track as Heldrup, also highlighting the challenges of opting out.

“What we see here in Sweden is that, first of all, you don't know how to make an opt-out. The AI office, which is the EU organ for AI regulation, has a recommendation that you should use a robot exclusion protocol to opt out from AI. However, when you use a robot exclusion protocol, you have to specify which crawlers you want to protect yourself from. We already have these AI companies coming up with new crawlers, and if you don't know which crawlers they are using, you're not able to opt out from those crawlers.”

Adding to the discussion, Norwegian Lisa Digernes raises another current concern: What about the material that has already been used to train AI, given that it's too late to opt out of the work that has already been scraped?

Copyright or fair use? – Legal approaches in the US, EU, and China

Digernes says the answer lies in the upcoming EU laws and the case development in the United States. There are several ongoing court cases in the US addressing AI and copyright law, for example The Authors Guild vs. OpenAI.

“There's no discussion on the substance yet, there are just procedural matters so far, so it's going to take at least three years before we see any decisions. But it's basically the authors suing OpenAI for copyright infringement on the training and on the output, and OpenAI says it's fair use. So this is kind of where we are.”

Fair use is a defence under copyright law, which under specific conditions allows limited use of copyrighted materials without the copyright holder’s permission. However, the legal landscape varies across jurisdictions, as Finnish Petteri Günther points out when comparing copyright laws in China, the US, and Europe.

“In China, my understanding is that the courts viewed the output as original because it was mainly determined by the choices of the user. In the US, however, when looking at the copyright office’s decisions, they haven’t yet found that prompting alone gives enough creative control over the output.”

Günther continues: “As for the EU, I think time will tell, but I wouldn’t rule out the possibility that, from a European law perspective, you could gain copyright for certain outputs, depending on how much you can influence what the final output looks like."

Are there legal frameworks that could change the current situation?

“I think the outcome of the lawsuits in both Europe and in the States are going to be the most interesting to see, the boundaries of the fair use doctrine, and for right holders to stand together to enforce basic principles of copyright. Because I think the DTM directive and innovation are obviously prioritised on behalf of the right holders and creative human productions.” Anna Stabell says.

She adds: “U.S. court, this fall, dismissed a lawsuit filed by news organizations over the unauthorized use of works for AI training, citing insufficient evidence of harm… This is not an encouraging development for how the legal system may assess the damage AI poses – and could continue to pose – to creators and copyright in general.”

Günther emphasises a cautious approach.

“I think at the moment there’s no obvious need to enact new legislation, and we should wait and see how the current laws work with the technologies we have now. The old rules apply, so the key factor is who makes the creative choices. That should be the starting point for discussions on generative AI today: Who controls the creation? Is there enough human control, or is it the machine that creates based on an idea you feed it?”

Hedvall, on the other hand, wants to see more radical changes, and advocates for a re-evaluation of the Text and Data Mining (TDM) exception in light of the Berne Convention. The TDM exception allows certain uses of copyrighted material for purposes such as research, education, or innovation, without obtaining prior permission from copyright holders. However, these exceptions must comply with the Berne Convention, which is an international agreement designed to protect creators’ rights over their literary and artistic works.

“The entertainment industry and the copyright industry are billion-dollar industries within the EU, and to basically allow these big AI companies to destroy the whole EU copyright system… I don't see how this is compatible with the three-step test in the Berne Convention, which is found in Article 9.”

To read the full Berne Convention: CLICK HERE.

Hedvall continues: “I think what we need to see in the EU is that the AI companies have to pay for the use of the creators' works. If you want to use the work of someone else, you have to pay them for it. I don't really see why the development of AI models should be free. Because they are charging their customers. There is money involved. Why should the creators not get a part of the cake, you know?”

Advice for creators in the age of AI

“I think, of course, you should experiment with these things, but you should be very mindful of what you put into a commercial product. Open AI recently released Sora, a video generating model. And you're basically able to generate content that looks like video games. So, you can produce a Mario game, a counter-strike game, and that's because they've trained on all this copyright-protected content. And there is a risk if you use these products right now that you will make something that is an infringement of somebody's work, and then you can be in a very bad situation at some point,” Thomas Heldrup says.

Lisa Digernes’ advice is to keep track of your AI use in the creative process.

“What we've seen so far is that owners of the tools are being sued, but you never know. But a piece of practical advice is to really keep track of what you're using AI for. Is it just for the idea phase, or are you using it to create final content that's going to go into your product? What type of prompts are you using?”

As many answers to legal questions surrounding AI are yet to come, Sigurður Örn Hilmarsson shares a thoughtful reflection on the current situation, drawing on the ancient Chinese proverb: "May you live in interesting times."

He notes: "That is actually a threat, because nobody wants to live in interesting times. We have to see this as a blissful glow for the benefit of all, or it will hit us like an avalanche."

For more conclusions, see the full discussion: