Wikimedia Draws Line Between AI Training, Output — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Wikimedia Foundation filed an amicus brief Friday in publisher lawsuits including the New York Times’ case against OpenAI and Microsoft, arguing both sides push “overbroad and simplified applications” of fair use.
- Wikimedia Foundation proposes that website scraping not count as copyright infringement and that unlicensed LLM pre-training qualify as fair use when models provide “something new and different” from source material.
- Wikimedia Foundation says outputs that copy protected work while supplanting readers’ engagement with the original should require a license, warning that black-and-white rulings could enable copyright gatekeeping or fragment knowledge flows.
- New York Times alleges OpenAI infringed its copyright by scraping the paper’s website to build training datasets; OpenAI and Microsoft contend that LLM training is fair use.
- Microsoft said in a September filing that expert analysis found Copilot rarely reproduces full sentences from news content used to ground the model, hardly undermining training’s transformative purpose.
- ChatGPT reproduces text freely licensed under Creative Commons without adhering to the license, according to Wikimedia, including Wikipedia’s attribution and same-license requirements for adaptations.
- Wikimedia Foundation identifies three ways LLMs can assist Wikipedia editors: fighting vandalism, detecting copyright violations and doxxing, and finding articles that need improvement.
Why it matters: For Wikipedia editors, the stakes are concrete: Wikimedia identifies three AI-supported jobs—fighting vandalism, detecting copyright violations and doxxing, and finding articles needing improvement—while saying ChatGPT violates Creative Commons attribution and remix conditions. The proposal puts licensing duties on AI companies and requires respect for Wikipedia’s contributors.
Ask SkimNews
