Wikimedia proposes a compromise position in AI copyright case: training may constitute fair use, but reproducing original text should require authorization
Wikimedia has proposed a compromise position in the AI copyright lawsuit, arguing that using copyrighted works without permission for large language model pre-training may qualify as fair use if the model's output differs from the original text, but that when chatbots directly output article content and replace readers' access to the original works, AI companies should comply with licensing requirements. Wikimedia points out that models reproduce Wikipedia text published under open licenses but do not necessarily comply with conditions such as attribution and share-alike, and it worries that if courts broadly rule that AI training is entirely unrestricted, this could open the door to ignoring licenses. The backdrop to this position is that The New York Times sued OpenAI and Microsoft in 2023 for scraping large numbers of news articles without permission to train AI systems, with the two companies arguing that such training constitutes fair use.
Why read this
AI score 58
Wikimedia has offered a layered compromise stance on training versus output in the AI copyright dispute, which can serve as a reference for understanding the direction of The New York Times v. OpenAI case.
Topics
Body · AI translation
Wikimedia believes that web scraping itself should not be regarded as copyright infringement; using copyrighted works without permission for large language model pre-training may qualify for the fair use doctrine if the model's output is "different" from the original text. But if a chatbot directly outputs article content, replacing readers' access to the original work, AI companies should comply with licensing requirements.
This argument is related to the Creative Commons Attribution-ShareAlike license used by Wikipedia. Wikimedia says models reproduce Wikipedia text published under open licenses but do not necessarily comply with conditions such as attribution and share-alike. The foundation worries that if courts broadly rule that AI training is categorically unrestricted, it could greenlight ignoring licenses and weakening the flow of content between authors and readers; conversely, if copyright holders gain overly broad control, it could also hinder innovative, transformative uses of works.
Wikimedia also points out that AI tools bring both risks and benefits: they can help Wikipedia editors identify vandalism, copyright violations and doxxing information, and find entries that need improvement; but chatbots providing information directly may also cause users to stop visiting Wikipedia and other original websites, weakening open knowledge projects that rely on community maintenance.
The New York Times sued OpenAI and Microsoft in 2023, alleging that the two companies scraped large numbers of news articles without permission and used them to train AI systems; both companies argue that the relevant training constitutes fair use. Wikimedia warns that if courts fail to carefully weigh the difference between training and generated content, rulings could affect future AI development, the open web, and how users obtain information.