Wikimedia joins NYT lawsuit to advocate for balanced AI copyright rules

A middle ground for copyright

The Wikimedia Foundation has officially entered the legal fray between The New York Times, OpenAI, and Microsoft. By filing an amicus brief, the nonprofit organization is asking a U.S. court to find a balanced path forward regarding AI and copyright. They are not taking a side against either the publishers or the technology developers.

Wikimedia expresses concern that a ruling that is too restrictive could harm the open internet and limit public access to reliable information. Conversely, they worry that an overly permissive ruling could damage the ecosystem of information sources. The foundation wants a decision that protects the integrity of the web while allowing for progress.

As Analytics Insight reported, the organization believes the court must distinguish between how AI models are trained and how they generate specific responses. This distinction is central to their argument about what constitutes fair use in the age of generative technology.

Training vs. retrieval models

The foundation makes a specific distinction between training large language models (LLMs) and using retrieval-augmented generation (RAG). Training involves feeding data into a system so it can learn patterns and create something new. Wikimedia argues this process should generally qualify as fair use if the output is different from the source.

RAG is a different technology that allows chatbots to pull information directly from websites to answer user questions in real time. Wikimedia argues that when AI systems use RAG to reproduce existing articles, they should respect the original licensing conditions. They believe companies should not automatically claim fair use for these types of direct retrievals.

This distinction is important because it separates the 'learning' phase of AI from the 'delivery' phase. By protecting the latter, Wikimedia hopes to ensure that publishers are compensated or credited when their specific content is served to users.

Protecting volunteer contributors

Wikipedia relies on a massive community of volunteers who maintain more than 67 million articles in 300 different languages. Much of this content is shared under the Creative Commons CC BY-SA 4.0 license. This specific license allows for reuse and adaptation, but it requires proper credit to the original contributors.

Wikimedia is concerned that if AI companies use this content without following these rules, the people who create the information will not be recognized. They argue that chatbots should be required to provide attribution when they use Wikipedia content. Without this, the labor of the volunteer community could be exploited without acknowledgment.

The foundation wants to ensure that the open nature of Wikipedia remains a benefit to the public rather than a free resource for commercial exploitation. They believe that respecting the terms of the CC BY-SA 4.0 license is essential for maintaining the health of the platform.

The stakes for the internet

The original lawsuit was filed by The New York Times in 2023, alleging that OpenAI and Microsoft infringed on copyrights by using published material. Both tech companies have defended their actions by claiming that using publicly available content for development falls under fair use. The court's decision will set a precedent for how all online information is processed.

Wikimedia warns that allowing unrestricted copying could weaken the viability of publishers and other information sources. If publishers cannot protect their work, the quality of information available on the web might decline. This could lead to a less reliable internet for everyone who relies on high-quality reporting.

On the other hand, the foundation notes that restricting AI training too heavily could slow down innovation. It could also limit the development of open-source technologies that many people rely on. They are seeking a path that allows for technological growth without destroying the underlying information economy.

What comes next for AI law

The court's upcoming decision will likely influence how AI companies collect, process, and present information in the coming years. Because Wikimedia is a major repository of human knowledge, their input carries significant weight in the conversation about data rights. The outcome will determine the rules of engagement for the next generation of software.

The foundation's stance suggests that the future of AI may involve a hybrid model of permissions. In this scenario, companies might be allowed to train on broad datasets while being required to pay or provide credit when they serve specific, high-value content. This would create a tiered system for how AI interacts with the web.

As the litigation continues, the focus will remain on where the line is drawn between inspiration and infringement. Wikimedia's involvement ensures that the perspective of open-source, community-driven knowledge is represented in the legal debate.