What potential legal issues can arise from incorporating generative AI in game development?
The gaming industry is poised to reap numerous benefits from the advent of generative artificial intelligence (GenAI).
This year, major developer events such as GDC have heavily featured companies showcasing GenAI technologies for creating multiple maps and levels, streamlining development workflows, conducting quality assurance, and interacting with player actions in real-time. Additionally, significant industry players like NVIDIA have introduced new hardware, such as the SUPER series of GPUs, to support these advancements.
Considering these developments, it's plausible to envision a future where GenAI significantly influences game development and gameplay. However, the excitement surrounding these advancements is tempered by the legal challenges that publishers must address before fully integrating GenAI into their processes.
The legal risks associated with GenAI span various laws, including intellectual property (IP), privacy, and tort/contract law. These concerns may arise either during the game's development phase ("DevAI") or during gameplay ("LiveAI"), and often affect both stages.
Intellectual Property Law
The integration of GenAI raises significant questions within intellectual property law, particularly around whether using training data that includes copyrighted content constitutes infringement. This issue is central to debates like the Palworld controversy, questioning if the GenAI's usage of copyrighted material inherently violates IP laws.
Avoiding IP Infringement based on AI Inputs and Outputs
This year's scrutiny over IP infringement involves notable cases like The New York Times suing Microsoft and OpenAI for using its articles to train ChatGPT's models. Microsoft's defense hinges on the concept of "fair use," which allows some use of copyrighted material without liability. The court ruling on this could set a precedent for training GenAI models.
Another crucial IP concern involves the outputs generated by GenAI models. For instance, if a GenAI model generates a town level featuring the Coca-Cola logo, it could be seen as trademark infringement. Similarly, creating a 3D character model identical to Mario would likely result in copyright issues with Nintendo. To mitigate such risks, developers should ensure that GenAI outputs are sufficiently original and not overly similar to existing protected content.
One measure to address this risk is to have human oversight or use another AI trained on a broader dataset to review outputs, ensuring they don't infringe on existing IP. However, these approaches aren't foolproof. Perhaps the safest strategy is to avoid using copyrighted content in training GenAI models in the first place. Even when models are trained on non-copyrighted data, IP issues can still arise, necessitating vigilant oversight.
Ensuring IP Protection of AI-Generated Content
Game companies usually seek to protect the content produced by DevAI, such as new weapon models, levels, or characters, through patents or copyrights. However, securing IP protection can be challenging when GenAI is involved. Notably, a court ruling in Thaler v. Vidal clarified that an AI cannot be credited as an inventor on a patent application, meaning GenAI alone cannot secure a patent for its creations.
Similarly, the U.S. Copyright Office has stated it will not register works created by machines without human intervention, emphasizing the need for human creative input. As per the U.S. Patent and Trademark Office's guidance, developers must play a significant role in directing or training the AI system to meet the requirements for patent protection. For copyright, it must be demonstrated that the works are the result of human creativity rather than mere mechanical reproduction.
Privacy Law
GenAI's integration also introduces privacy law challenges, particularly concerning the collection and use of personal data. States like California have stringent consumer privacy laws, such as the California Consumer Privacy Act, regulating how consumer data is handled.
In scenarios involving LiveAI, and potentially DevAI, users' personal information could be used to train GenAI models either directly or indirectly. Sharing these AI models or their outputs might inadvertently disseminate personal data, implicating privacy laws. Developers must be cautious about how they manage and share personal data to comply with these regulations and avoid legal repercussions.
Tort Law/Contract Law
Tort law grants individuals the "right of publicity," allowing them to control the commercial use of their identity. A breach of these rights, such as using someone's name or likeness without permission, can result in legal action. For example, if GenAI trained on a player's facial expressions later reproduces their likeness in a game, this may constitute a violation of their right of publicity.
To mitigate these risks, game developers should include disclaimers in terms of service agreements, informing players that their data may be utilized in the game's development or future projects. However, this alone may not suffice, as overbroad licenses hidden in legal jargon could result in negative public relations.
A critical concern involves voice and motion capture actors, highlighted by recent SAG-AFTRA strikes. Actors are wary of their likenesses being used in GenAI models, potentially reducing their future job opportunities. Consequently, developers need to craft contracts carefully to respect actors' rights and consider alternate training methods to prevent unintentional recreation of an actor's likeness.
Further, the proposed NO FAKES Act aims to protect individuals from unauthorized AI-generated reproductions of their likeness, adding another layer of complexity for developers using GenAI.
The potential for legal complications underscores the need for robust internal policies to regulate AI use. Developers and publishers must remain vigilant and proactive to avoid these pitfalls. If these precautions are taken, game developers can safely harness the power of GenAI, ensuring successful and lawful integration, as famously expressed by an iconic AI from gaming, "This was a triumph. I'm making a note here: huge success."
Andrew Velzen, an intellectual property expert at MBHB, advises clients on IP matters involving technologies like machine learning and artificial intelligence.
