The Wikimedia Foundation’s Stance in Ongoing AI Copyright Lawsuits

The Wikimedia Foundation is positioning itself in the midst of significant copyright litigation regarding the use of artificial intelligence. The organization is advocating for legal protections surrounding large language models (LLMs) that are trained on copyrighted materials, while simultaneously warning against permitting chatbots to simply replicate this content for users. A blanket ruling, it argues, could either restrict the beneficial use of AI tools for content creators or jeopardize the integrity of the open web.

In an amicus brief submitted last Friday, the nonprofit organization, which manages Wikipedia, refrained from allying with either side in the disputes involving major publishers like The New York Times against entities such as OpenAI and Microsoft. The brief states, “Wikimedia is concerned that the parties’ positions in this case push for overbroad and simplified applications of the fair use doctrine” regarding AI technologies. The Foundation warns that a simplistic victory for publishers might lead to copyright holders exerting control over transformative and innovative uses of their works. Conversely, a win for AI companies could create a precedent allowing them to bypass licensing agreements, disrupting the knowledge flow between authors and their audiences.

“Wikimedia is concerned that the parties’ positions in this case push for overbroad and simplified applications of the fair use doctrine.”

This ruling carries significant weight for Wikipedia, as the Foundation notes that its platforms have evolved into vital resources for training various AI and machine learning systems. However, there is concern that these AI systems are redirecting web traffic away from Wikipedia, threatening the volunteer-driven community that sustains it. Additionally, Wikimedia claims these tools are infringing copyright; many services, like ChatGPT, reproduce text from its Creative Commons-licensed materials but fail to comply with the license’s stipulations.

While acknowledging the challenges, Wikimedia recognizes that LLMs offer valuable functionalities to Wikipedia, such as assisting editors in combating vandalism, identifying copyright infringements, and improving article quality.

“A ruling that does not apply fair use principles carefully could threaten the sustainability of free knowledge.”

According to Stephen LaPorte, General Counsel at the Wikimedia Foundation, the courts must appreciate that Wikipedia thrives on a collaborative model grounded in the free access to human knowledge. He elaborated that the implications necessitate a careful review to prevent stifling the development of open-source technologies, which are crucial for broadening information access.

The New York Times advocates for a broader interpretation of fair use in relation to LLMs, alleging that OpenAI infringed its copyright by using its website to develop training datasets. While both OpenAI and Microsoft maintain that their utilization of LLMs constitutes fair use, OpenAI claims that replicating original material is an infrequent error. Microsoft argues that expert evaluations show that its tool, Copilot, seldom reproduces entire sentences from sources, asserting it rarely undermines the transformative purpose of LLM training.

Wikimedia’s proposal suggests a balanced approach: it argues that scraping content from websites should not be classified as copyright infringement. Instead, it envisions that using copyrighted materials to pre-train LLMs should be recognized as fair use, provided they yield “new and different” outputs. However, delivering copies of material in direct response to user inquiries, thereby displacing user engagement with the original sources, should necessitate a license. In such situations, AI companies must adhere to licensing requirements, ensuring that original sources receive due credit under Creative Commons licenses, including in any remixes or adaptations.

The CEO of Wikimedia, Bernadette Meehan, stated, “We are taking a position in this case to defend the free and open human knowledge that Wikipedia depends on.” She emphasized that while the expansion of knowledge access through AI is a noble objective, it must honor the contributors and communities that ensure the reliability of knowledge.