AI Fan-Made Games and Copyright Law: Creativity, Challenges, and the Future of Open-Source Game Development

AI Fan-Made Games Using Open-Source AI: Unlocking Creative Power

Fan-made games leveraging open-source AI models are opening new doors for creativity in game development. However, legal frameworks and copyright laws pose significant challenges due to the unauthorized use of data. Developers must consider safeguards like creating custom models, maintaining detailed documentation, and establishing clear disclosure policies to address these issues effectively.

Fan Communities Leading AI Innovation

While traditional media production has been cautious in adopting generative AI, fan communities have quickly embraced open-source AI models and interface frameworks such as Ollama. These enthusiasts create innovative applications, including chatbots and story-driven games based on intellectual properties. This approach highlights the evolving participatory culture, where human creativity intertwines with AI and data in complex ways.

Media scholar Nicole Lamerichs notes that this intersection involves humans, nonhumans, businesses, data, and interfaces interacting in new forms. Yet, the legal challenges remain centered on copyright issues and questions about whether AI-generated content can be protected.

An Example: Stranger Things Fan-Made Text Adventure

A noteworthy case is a fan-made Stranger Things text adventure game built with off-the-shelf AI tools. The developer scraped approximately 8,000 words from the fan-created Stranger Things Wiki and unauthorized scripts for character dialogue. The content is organized by key locations from Seasons 1 and 2, such as Hawkins Middle School and Mike’s neighborhood.

Players navigate these locations by typing simple commands like “East” or “West,” reminiscent of classic text adventures, while staying true to the 1980s setting of the show.

The game runs on large language models (LLMs) hosted on a local server with a powerful GPU, enabling near-instant text generation. However, this hardware requirement limits accessibility for most users. The developer experiments with various LLMs available through Ollama, including smaller models like Llama and larger ones like DeepSeek R1 and Qwen 3, to compare their performance.

Innovations in Game Mechanics Using AI

LLMs drive more than just narrative generation. For example, one model evaluates battle outcomes against fan-created bosses such as the ‘AquaGorgon,’ while a creative writing model delivers rich descriptions of the battle results. This setup allows the game to track win/loss stats and provide an engaging storytelling experience.

Another inventive feature is AI-generated in-game items. These are created as text strings by the AI rather than hard-coded assets. The same AI evaluates their usefulness during battles, blurring the line between game data and creative writing. This method opens up endless possibilities for content and development workflows.

Aligning with the show's 1980s tabletop RPG vibe, instructing the AI to act as a Dungeons & Dragons game master helps it better understand its role.

Legal Challenges: Copyright and AI-Generated Content

Despite the creative potential, copyright infringement remains a major concern for AI-powered fan works. AI models are often trained on vast amounts of internet data without explicit permission from rights holders, raising questions about unauthorized use.

Numerous class-action lawsuits have been filed against companies like Microsoft, OpenAI, and Stability AI over copyrighted material in training datasets. In the U.S., “fair use” has offered some defense, such as in Google's case against Oracle, but Europe lacks an equivalent doctrine.

If an AI model trained on copyrighted content produces outputs closely resembling existing works, using those outputs in games could infringe copyright. Platforms like Steam have started rejecting AI-generated games unless developers can prove they have rights to all intellectual property involved in training their AI.

This situation highlights the unclear legal ownership of AI-produced content. Using AI models trained on copyrighted game assets or storylines to create new mods risks violating developers’ rights. At the same time, copyright protection of AI-generated content remains an unresolved issue for game creators.

Insights from Legal Cases on AI Copyright

Generally, copyright law in the U.S. and Europe grants protection only to works created by natural persons, excluding AI as an author. For example, in Thaler v. Perlmutter, the U.S. Copyright Office denied registration since the author was a computer system.

A key requirement for copyright is a “threshold of originality,” meaning the work must be independently created and show human creativity. The extent of human involvement is crucial.

In the case of Zarya of the Dawn, the U.S. Copyright Office granted copyright for the comic book except for AI-generated images, ruling that the AI's contribution was too automatic and random. Minor human edits to those images did not qualify for protection.

Conversely, China’s Li v. Liu case recognized copyright for an AI-generated image because the plaintiff demonstrated precise human direction over the AI.

While the U.S. Copyright Office suggests some AI-assisted works might qualify for registration if they reflect original human mental conception, pure AI-generated works face significant hurdles. Developers may struggle to prevent others from using AI-created characters without substantial human authorship.

Practical Steps for Developers

• Choose AI tools that allow full user control over the generated content and clarify ownership rights.
• Read and understand the terms of service for AI tools, especially regarding content ownership and usage rights.
• Use custom AI models trained on proprietary or licensed data whenever possible.
• Customize AI-generated outputs extensively to ensure they serve as foundations rather than finished assets.
• Keep detailed records of human input, like prompt history and editing, to demonstrate creative involvement.
• Regularly audit AI content for potential copyright issues.
• Establish internal policies about AI use and require disclosure when submitting games to platforms like Steam.

These measures help developers reduce legal risks while experimenting with AI-driven content.

The Road Ahead

The fusion of AI and fan-made games offers exciting creative opportunities but also introduces legal uncertainties. The blurred boundary between code and creative writing enabled by LLMs creates both innovation and challenges.

For now, fan games built on open-source AI exist in a gray area where legal frameworks have yet to catch up. Changes in copyright law and regulation will be necessary to support creativity while protecting original works.

Until then, developers must balance innovation with caution, adopting best practices to respect intellectual property rights and document their creative process thoroughly.