The short version
- The Wikimedia Foundation filed an amicus brief advocating for a distinction between training AI models on copyrighted data and generating outputs that replace original works.
- Officials warn that overly broad rulings could either allow copyright holders to block transformative uses or enable AI companies to ignore licensing terms, threatening the sustainability of free knowledge platforms.
- The organization emphasizes that while LLMs offer valuable tools for maintaining Wikipedia, they must adhere to Creative Commons licenses when reproducing content for users.
The nonprofit organization behind Wikipedia has entered a complex legal debate over artificial intelligence and copyright, urging federal courts to adopt a nuanced stance rather than issuing sweeping rulings that could harm the open web. In an amicus brief filed recently, the Wikimedia Foundation positioned itself between publishers seeking stricter protections for their content and technology companies defending the use of large language models. The group argues that current legal arguments from both sides are too simplistic and risk creating outcomes that damage the ecosystem of free information.
At the heart of the dispute is the question of how fair use applies to AI systems. Publishers, including The New York Times, contend that scraping their websites to build training datasets constitutes copyright infringement. They argue for a more expansive definition of fair use that limits how companies like OpenAI and Microsoft can utilize protected material without permission. Conversely, these technology firms maintain that training models on such data is a transformative use that falls squarely within fair use protections, asserting that the resulting AI systems do not merely copy content but create new value.
Wikimedia rejects this binary framing. The foundation warns that a decisive victory for publishers could empower copyright holders to gatekeep even innovative and transformative applications of their work. Such an outcome might deprive Wikipedia editors of useful AI-based tools that help combat vandalism, detect copyright violations, identify doxxing attempts, and locate articles needing improvement. These technologies have become integral to maintaining the quality and integrity of the encyclopedia, which relies on a collaborative process supported by accessible human knowledge.
On the other hand, the foundation cautions that a complete win for AI companies could signal permission to disregard licensing terms entirely. This scenario might fragment the flow of knowledge between human authors and audiences, undermining the communities that sustain free information platforms. Wikimedia notes that its websites have become critical data sources for machine learning training. However, these same systems often siphon away visitors, reducing engagement with the original sites and threatening the volunteer networks that keep them running.
The organization proposes a compromise centered on the distinction between training and output. According to Wikimedia, scraping websites should not be considered copyright infringement, and using copyrighted works to pre-train large language models without a license should qualify as fair use if the result offers something new and different from the original sources. This approach acknowledges the transformative nature of AI development while respecting the foundational role of existing content in building these systems.
However, the foundation draws a firm line when it comes to AI-generated responses. If models reproduce copies of material in response to user queries and deliver them in a way that supplants readers’ engagement with original works, such actions should require licensing. In these instances, the output is merely substituting for the original rather than transforming it sufficiently to merit fair use protection. This distinction aims to prevent AI services from bypassing the economic and creative incentives that copyright law is designed to protect.
For Wikipedia specifically, this means that any AI service utilizing its content must adhere to Creative Commons licensing requirements. These licenses mandate crediting the original source and ensuring that any remixes or adaptations remain under the same license terms. Wikimedia argues that services like ChatGPT have reproduced text licensed freely but failed to comply with these conditions, thereby violating the spirit and letter of the agreement. Such violations threaten the sustainability of free knowledge by eroding trust in the licensing framework.
Stephen LaPorte, general counsel for the Wikimedia Foundation, emphasized that a ruling failing to apply fair use principles carefully could stifle open-source technology development and limit access to information for everyone. He stressed the need for courts to recognize the collaborative nature of Wikipedia and its dependence on a strong foundation of open knowledge. Bernadette Meehan, CEO of Wikimedia, added that expanding access through AI is a powerful goal but must respect the people and communities that make reliable knowledge possible.
The case highlights broader tensions in the information ecosystem as AI technology continues to evolve. Finding a balance between protecting creators and fostering innovation remains a shared responsibility for all stakeholders involved. As courts deliberate on these landmark cases, their decisions will likely shape not only the future of artificial intelligence but also the viability of open-web platforms that rely on voluntary contributions and free access. The outcome could determine whether the digital landscape becomes more fragmented or remains an interconnected resource for global knowledge sharing.
Legal experts and industry observers are watching closely to see how judges interpret these competing interests. The stakes extend beyond individual companies or publications, affecting the broader principles of intellectual property in the digital age. If courts adopt Wikimedia’s proposed middle ground, it could establish a precedent that encourages responsible AI development while preserving the integrity of open knowledge systems. Conversely, a failure to find common ground may lead to prolonged litigation and uncertainty for both creators and technologists.
Sources behind this briefing
Go to the original reporting
- The Verge↗Wikipedia’s operator says a bad AI copyright ruling could threaten the open web
