Game studios have a lot to navigate when it comes to AI use. Generative AI is increasingly becoming embedded in the game development process, and at the same time, the laws, regulations, and guidelines surrounding it are on the rise. Each landmark case brings with it new nuances around how AI should and could be adopted.
So alongside grappling with the general, broader uptick in games regulation (see this article for more on how recent laws affect game design), studios also have to contend with the specific legal risks around the use of generative AI and its changing laws. These changes run parallel to any requirements a studio might have with its partners and any AI policies that they might require
Knowing how to address the myriad new legal considerations and compliance requirements can understandably seem a daunting task. So here is a high-level run through of some of the key areas in which the changing AI legal landscape is affecting game studios.
The IP infringement risk of LLMs
The nature of LLM training (namely the ingestion of data sets containing a wide array of third party works, in many cases without consent) creates a risk that AI-generated output could occasionally contain infringing material. This is setting aside the fiddly issue of potential infringement at the training level.
There is another potential risk in that many AI tools require that any works inputted or generated by a third party AI tool are licensed back to that tool provider. In other words, while a studio might own an output, this output is then licensed back to the tool provider for further use. If the input or output included proprietary IP, then this means that proprietary IP could be used for, among other things, AI training purposes.
Some AI tools address this issue, especially in their premium offerings, by making it a one-way process, so nothing goes back to training the model – but this needs to be confirmed and checked.
Studios should therefore implement clearance processes to ensure that any prompts and any outputs are carefully vetted. For instance they should check for riskier inputs (such as those specifically targeting third party IP, or any that might include confidential information or personal data) and potentially infringing outputs (for example, a background asset that features a third party’s logo or a generated human face that bears a striking similarity to a real celebrity).
These clearance processes need to be aligned with any other requirements that a studio might have. For example, if the studio is working on a project for a publisher, then the publisher might require additional AI safeguards.
Rules around AI use and performers have tightened
The long-running SAG-AFTRA strike of 2024/25 culminated in an agreement placing restrictions around how AI is used with a performer’s voice or likeness (including requirements for express consent and compensation). This is now increasingly the market baseline for actors’ contracts, even if the performers are not union members.
The data privacy risk of AI
The prompts entered into third-party AI tools, along with their outputs, are often not as confidential as many may think. Even with a private or enterprise-tier subscription, the inputs and outputs are often still accessible by the tool provider, for example to provide backend support, and at times more broadly for general internal purposes such as AI training.
This means that studios need to be careful when it comes to any confidential information or personal data which may be included in the prompts, and they should also be careful when it comes to generating works that need to be kept confidential.
Who owns AI outputs?
On one level, ownership of the output generated by an AI tool will depend on the relevant terms and conditions for that tool. But the ownership of AI-generated works is also subject to the laws of the relevant jurisdiction, which are evolving in different ways from country to country.
In the US, for example, you cannot own the copyright for an image that is solely AI-generated. But you can own AI-generated works to the extent that they have been modified by a human, even if you do not own the underlying, unedited image.
By contrast, in other jurisdictions such as China, AI-generated works have been granted copyright protection. In the UK, the IP for an AI-generated work is owned by the person who made the necessary arrangements for the creation of the work (provided that the work is sufficiently original).
Given the shifting legal landscape across different jurisdictions, studios should have meaningful human involvement in its AI processes and, with respect to important external-facing works in particular (where ownership is a particular concern), a more substantial level of human editing is a must.
Advertising and AI
The UK’s Advertising Standards Authority (ASA) has been releasing guidelines regarding AI content and adverts. The overarching message remains the same as it has always been: nothing misleading, and nothing degrading or demeaning. This is nothing new when it comes to advertising, but particular attention needs to be applied to realistic AI-generated content and AI assistants.
For example, the ASA has ruled against certain adverts for apps in which users can create AI characters that can engage in sexually suggestive chats, or that can be used for other sexualised simulations, on the grounds of objectification of women.
Real-world settings come under regulation
As Antonia Gold explained earlier this year, the definition of “deepfake” under the new European Union AI Act, which came into force in August, is broader than you might think.
The AI Act defines a deepfake as AI-generated or AI-manipulated content that resembles existing persons, objects, places, entities, or events and would falsely appear to be authentic or truthful. This means that if a studio has used AI to create any content that might create a false impression of authenticity – for example, celebrity voices, or real events or people depicted with a high degree of photorealism – then they will probably need to label that content as AI-generated.
When does AI use need to be declared?
In the midst of all this legal moving and shaking, storefronts such as Steam have taken matters into their own hands and implemented various rules regarding the disclosure of AI-generated content.
Confusingly for games businesses, each platform has its own rules regarding AI labelling and disclosure (and some have none at all), meaning that studios need to assess in each case what a platform’s requirements are on top of any legislative obligations – such as potential labelling requirements under the EU AI Act.
Contractual obligations
Protections around AI use have become a new contractual staple across the industry, and can govern anything from which AI tools are used and how to what constitutes your business’s “AI hygiene” and responsible use of AI.
Whether in contractor agreements, talent agreements, investor terms or EULAs, AI provisions in their various forms are being implemented as standard.
The upshot: compliance, compliance, compliance
All this boils down to an uptick in studios’ internal compliance requirements. Having an AI use policy is now no longer a “nice-to-have”, but a “have-to-have”, and can be something a studio agreed to in its contracts with its partners.
Investors, publishers, insurers, and auditors are increasingly requiring this. More than simply being a box-ticking exercise, studios need to have a way of ensuring a business-wide understanding of potential legal issues and how to mitigate the risks of them. Once a policy is in place, it should not be put in a drawer never to be looked at again, but should instead be the start of a practical rollout across the business, which in some cases can prove to be a project of mighty proportions.
The games industry is already well-adapted to thinking creatively in the face of new technology
In short, the legal landscape surrounding AI use in game development is evolving at a pace that can feel hard for studios to keep abreast of.
The games industry is not alone in this. Other creative industries have also been looking at ways of navigating widespread AI adoption against the backdrop of these legal uncertainties.
But the games industry has a key advantage over other sectors: it is an industry that is already well-adapted to thinking creatively in the face of new technology.
The rapidly evolving AI legal landscape does not mean that studios cannot leverage AI technology. Rather, they need to have guardrails in place for using it lawfully, because once they can pinpoint where the legal risks are and how to work around them, they can then build their own way of using this technology safely, responsibly, and creatively.
Anna Poulter-Jones is a partner in Sheridans’ Games team. She advises games studios of all sizes on a variety of commercial, regulatory, and intellectual property matters. She is also a board trustee of the National Videogame Museum and a member of the Appeals Panel for the Games Rating Authority.
UKIE has been working with Sheridans to roll out an AI Compliance and Accreditation Service. It will serve as an industry-wide AI compliance framework, open to all games studios, to provide a consistent standard for responsible and lawful AI use across UK games, that is still tailored to each studio. For more information, please visit: https://ukie.org.uk/ai-certification-mark/.