Wikimedia draws a copyright distinction between training and grounded answers

Wikimedia’s [October 30 statement](https://wikimediafoundation.org/news/2026/10/30/wikimedia‑foundation‑amicus‑brief‑new‑york‑times‑v‑open‑ai‑case/) argues for fair use when training produces models that are new and different from their sources. It argues that grounding which delivers copied source expression should respect publisher licences. This is the Foundation’s argument, not a court ruling.

The Foundation also says scraping should respect public API rate limits or use its Enterprise offerings, regardless of purpose. Permission to access content and the copyright treatment of its use answer different questions. Lady Justice, Melbourne, October 2023; contextual photograph, not the New York case.