The AI state of play: AI laws and game development

Generative AI is now embedded in game development, but evolving laws create new legal risks around training data, outputs, and third-party tool licensing. Studios must vet prompts and outputs to avoid infringing material or proprietary IP being licensed back to providers. Clearance processes are essential, especially when working with publishers or handling confidential data, as AI tools often retain inputs and outputs for training or support.

The 2024/25 SAG-AFTRA agreement restricts AI use of performers' voices or likenesses, requiring express consent and compensation, now a market baseline even for non-union actors. Confidentiality risks persist, as third-party AI tools may access inputs and outputs for backend support or broader internal purposes, including training. Studios must treat AI-generated works with caution, considering both tool terms and jurisdiction-specific laws on ownership.

The UK’s Advertising Standards Authority (ASA) has issued guidelines targeting misleading or degrading AI-generated content, particularly in ads featuring realistic AI characters or simulations. The EU AI Act, effective August 2024, broadens the definition of deepfakes to include AI-generated content that could falsely appear authentic, requiring labeling for such material. Platforms like Steam have also introduced AI disclosure rules, adding to the complexity of compliance for studios.

AI provisions are now standard in industry contracts, covering tool usage, responsible AI practices, and internal compliance policies. Studios must implement AI use policies to meet investor, publisher, and insurer requirements, ensuring business-wide understanding of legal risks and mitigation strategies. These policies should be actively enforced rather than treated as static documents.