Source: Wikimedia Foundation (Diff)

The Foundation says the rapid rise in generative AI use and the corresponding surge in demand for continuous training data threaten the social contract that sustained the free knowledge movement. It reports that automated web crawling and scraping plus opaque model training practices have changed how people obtain and consume knowledge, and that these shifts risk severing the link between human contributors and those who benefit from their work. The Foundation highlights that Wikimedia projects mostly use Creative Commons licenses—primarily CC BY‑SA 4.0—which historically ensured attribution and ShareAlike reciprocity, but that current practices risk undermining those norms.

The post identifies three concrete threats: (1) inadequate transparency and bookkeeping about what goes into AI training datasets; (2) AI outputs frequently omit or misalign citations, though retrieval‑augmented generation (RAG) could enable better source citation; and (3) AI interfaces often prevent users from discovering and contributing back to the free knowledge ecosystem, while AI‑generated content poses editorial burdens. The Foundation notes ongoing experiments—model cards, an Attribution API and framework—and states a commitment in its annual plan to review licenses and legal policies to protect and sustain the knowledge commons and to propose specific actions in coming months.