“At some point AI will become the norm, and the law will catch up”
NORDIA Law's Tuomas Pelkonen reveals how AI will reshape legal practices and advises developers on the risks.
Welcome to this week’s edition of the AI|G newsletter. Each week, we interview an expert from the games industry about the role of artificial intelligence. We’re building a fine collection of Q&As with people from all disciplines, including marketing, development, monetisation, UA and more, so please subscribe for free and encourage your colleagues to do the same. This time, it’s the turn of Tuomas Pelkonen, Senior Associate/Attorney at NORDIA Law, who shares his insights into how the law views AI and game development. Pelkonen has a background as a competitive gamer, too, so he’s uniquely placed to talk about the intersection of games, law and AI.
Tuomas Pelkonen, NORDIA Law
In this week's Q&A, we sit down in Helsinki with legal expert Tuomas Pelkonen. We discuss how AI impacts intellectual property rights, the copyright status of AI-generated content, the potential risks and benefits of using AI in the workplace, and the evolving legal landscape. We also touch on the future implications of AI for games and consider how to approach AI integration responsibly. One fascinating nugget is how the ability of AI to analyse and summarise documents quickly could impact the way lawyers bill in future.
The top takeaways from our chat are:
Currently, AI-generated content generally isn't eligible for copyright protection in the US and Europe, but there are nuances based on the AI tool used and the level of human creative input.
To mitigate risk, when using AI, developers should use AI tools that offer maximum user control over output. Review the terms of use carefully and maintain a clear AI usage policy.
Unlike some tech trends, AI is likely to become a permanent fixture in game development, but adoption will be gradual due to legal considerations. Companies need to balance the risk of falling behind against the risk of legal issues.
AI Gamechangers: Tell us about your personal journey, please. You started as a competitive gamer before working in the law?
Pelkonen: I played EA Sports NHL since I was a kid. I’ve been playing NHL for about 30 years! When the competitive gaming scene picked up around six years ago, I started competing more actively, and I got pretty good, finishing among the top 32 best players in Europe in the NHL World Championships. That was last year, which was a surprise for me because I don’t have much time to play compared to when I was a student. The game hasn’t changed that much, though, and it’s about how you master the mechanics and how you handle the pressure.
When I graduated I wanted to work at a law firm, and business law especially. I applied to a few law firms, and NORDIA just felt right. I didn’t have a clue that they were working so much with games companies! It was a major surprise for me when I received an assignment and the client was Remedy Entertainment. I said, “Wow, I have been playing their game since I was a kid!” That was like a dream come true.
Since then, I have worked with many gaming companies, and I’ve started to build our gaming practice even further.
Is there much AI in your background? How did you come to have AI in your field of legal interest today?
I have a very broad educational background. I’m also an economist. And I’m interested in technology, and I’ve been following the conversation around AI for a while. When ChatGPT was released to the public, I quickly noticed that it’s a game-changer.
“It’s definitely safer to not use AI at all! But if you do not use it at all, you may fall behind. At some point, I believe, it will become the norm, and the law will catch up.”
Tuomas Pelkonen
The law industry has traditionally been very conservative. However, I can already see multiple use cases where we can utilise AI, and I think it can change the industry. Look at where ChatGPT, Copilot and other AI tools currently excel, for example. They analyse and summarise text really well. Let’s say you are working for a potential buyer and doing due diligence. Inspecting the company they are planning to purchase, you find, let’s say, 1000 agreements, and you have to check all the change of control clauses in case the transaction will affect these agreements. This transaction may trigger this change of control clause, and then the other party may have a right to terminate the agreement. If you have to find all of those clauses from 1000 agreements, those AI tools can do it very quickly! It’s easy then to verify those findings.
I think it might even change the industry in that way, because lawyers bill by the hour. In the future, it may be that we save so much time using AI that billing will be more based on value provided for the client rather than the time spent.
What are the most pressing concerns that game developers should be aware of when they start using AI in the workplace?
There is a concern that copyrighted material is used as training data for AI tools. When that’s the case, the output may also reflect that. So the worst case is that the AI tool just generates a copy of some existing work, and if you don’t recognise that, you may end up using it in a game. Not only might it be very difficult to take out later, it can lead to a dispute.
Is it clear at the moment where copyright lies with something generated by AI?
It’s an excellent question, because it’s not clear. But the current legal consensus, especially in the US and Europe, is that, in general, AI-generated content is not eligible for copyright protection because it lacks human authorship. There isn’t usually a sufficient amount of human intervention and creativity used in the process.
This is the current consensus. However, the US Copyright Office has given some guidance related to copyright registration; they state that it’s possible to get your AI-generated content registered, depending on two factors. It has to be analysed case by case, but those decisive factors are what AI tool is used and the second thing is how you actually use it. It becomes about the prompts, how detailed they are, and how well they reflect your creative process and thoughts. The end result will have to be that there’s a sufficient amount of human intervention and creativity used, and then it may receive copyright.
But consider a tool like Midjourney, for instance: that’s one that the US Copyright Office has said seems to work automatically and randomly. It generates four images based on your prompts – or from an image you provide from the internet – and then you can select one, and ask for a new variation on that, or to enhance its resolution, or just ask for a new image. Every time you get something new and it does most of the thinking.
If a game developer wants to start using AI, is there anything they should be aware of?
First, pay attention to the AI tools you select. Select the one which gives you, the user, a maximum level of control regarding the output.
Then it’s important to review the terms of use. Pay attention to who owns that output – it may be that the ownership is shared. Make sure that you, as a game developer, get all the rights to the content. Also, pay attention to whether you have to provide a license to the AI to use your output, such that it gives them the possibility to use that material as training material.
People are worried about their jobs. What’s your perspective on the impact AI will have on the job market?
AI can automate tasks. It speeds up your work a lot, but I don’t see that it replaces anyone easily. Maybe some very routine tasks can be replaced by AI, but I don’t think many people’s jobs are about just routine tasks!
It’s a good servant, but it’s a bad master. It cannot do all the work for you. And, of course, you need to check its work. To give you a legal example, you’ll have heard stories that lawyers asked ChatGPT to provide some case law, and the case law provided was simply made up. You need to utilise it correctly.
How has the legal landscape evolved to take AI into account?
The fact is that the law always struggles to keep up with technology. That has always been the case, so nothing new here.
There are issues that we have to address. Do you get copyright protection for your work, for instance, or does the training of AI infringe copyrights (if protected work was used for training material)? It takes many years to draft a new law and get it approved, and usually many years to resolve court cases. We already got the AI Act, but it doesn’t give all the answers. We just have to address issues by applying existing laws and regulations. It may lead us to some difficult situations where we don’t have the 100% correct answer to each question.
We draft AI policies for companies to make sure everybody is on the same page regarding the use of AI. It’s obviously very important that you can protect your IP. You also have to disclose the use of AI. For example, Valve requires game companies to disclose the use of AI [on Steam]. I can already see that if there is an M&A transaction, the buyer will want to see the AI policy. It’s already often asked in a due diligence process.
It’s definitely safer to not use it at all! But if you do not use it at all, you may fall behind. At some point, I believe, it will become the norm, and the law will catch up. Think of it: one of the main purposes of copyright and patent laws is to incentivise people to make creative works and inventions. I think at some point the law will have to make it possible that you can use AI safely and get your work protected. But it’s very much balancing between protecting existing works and also making it possible to get copyright for new works that are created with the use of AI.
Please think ahead and give us a prediction: just how disruptive is AI going to be for the games industry?
Consider crypto and NFTs. People were thinking that these would change the industry a lot… but then the bubble kind of burst, and maybe it didn’t become such a big thing.
“You should be aware that when copyright material is used as training material, the output may reflect that and be a direct copy of some existing work. There’s an argument that this was fair use under the Fair Use doctrine… In Europe, we don’t have such a legal framework”
Tuomas Pelkonen
But with AI, I think it’s going to be a longer-term thing. It’s going to stay and be the norm. I think that’s going to take a while because companies have to be a bit careful because there are these issues related to the use of AI. It will take several years to become a norm, but I think it’s definitely here to stay.
As a gamer, what excites you the most about AI technology?
All the possibilities and how easy everything is. I could make a game using AI without any knowledge of programming.
This all fits with user-generated content because it makes it easier for users to generate content with the help of AI. It could really boost that and make it an even bigger thing. Look at all the mods, such as users have done with GTA – those are just amazing to see!
What is your advice for indie teams who maybe can’t afford a lawyer? How should smaller teams approach AI?
All the advice here goes for them too. I think, use AI at the beginning and maybe in the middle phase of your creative process. Do not make something where you use the output directly in a game. Modify it, enhance it, use it as a starting point, so that reduces your risks considerably.
You gave a talk at Pocket Gamer Connects Helsinki last week about AI and the law. Can you give us an overview of your presentation?
I present a few court cases related to using copyright-protected material as training material. The learnings from these cases is that you should be aware that when copyright material is used as training material, the output may reflect that and be a direct copy of some existing work. Only time will tell what the courts will decide about these cases. There’s an argument that this was fair use under the Fair Use doctrine. What’s interesting is that in Europe, we don’t have such a legal framework; there’s no such doctrine, so even if they might win in the US, we will then have to assess the situation in Europe again, applying the regulation here.
I present a case from the United States where an author was not granted copyright for her comic book. She used Midjourney to generate the images for a comic book – she got all the copyright to everything else, but not the images that were generated using Midjourney. I go through why that was the case.
Then, I have another example where a court ended up with a totally different decision. This one was in China. The plaintiff actually got the copyright for the images he generated using Stable Diffusion. It’s interesting that the arguments were pretty similar; each was trying to convince the court that they used their creativity, and the end result just gave additional expression to their creative work. But this court said, “Okay, you really were able to direct the AI accurately”, and he got the copyright.
What differs was the AI tool used, Midjourney in the US but Stable Diffusion in the China case. So you definitely need to pay attention to which AI tool you use!
Further down the rabbit hole
Some useful news, views and links to keep you going until next time…
Filipino startup Clout Kitchen raised $4.45m towards its League Of Legends “AI buddies” platform, Backseat AI.
AI featured heavily in the Noble Prize line-up this year. The physics prize went to AI pioneers John Hopfield and Geoffrey Hinton. The chemistry prize was split between the University of Washington’s David Baker, and Demis Hassabis and John Jumper of Google DeepMind. Demis Hassabis is a former game developer.
NVIDIA released its first GeForce graphics card exactly 25 years ago. The GeForce 256, released in 1999, was the world's first GPU and not only revolutionised graphics performance in games but also laid the foundation for advancements in AI tech.
There are AI talks and panels scheduled for the upcoming Pocket Gamer Connects Jordan and London conferences. Tickets are still available for both events.
Third Dimension has secured $7 million in funding to develop generative AI technology that can rapidly create 3D game worlds from 2D inputs. Led by CEO Tolga Kart, who has experience from Activision and Tesla, the startup aims to dramatically reduce game development time and costs.
