Wikimedia Deutschland (the German chapter of Wikimedia) announced on October 1, 2025, the launch of the "Wiki-LLM" project, aimed at systematizing and simplifying access to Wikipedia data for AI developers. Although Wikipedia is one of the primary sources for training large language models, its standard data "dumps" are difficult to process due to specific markup and a lack of unified structure. The new project, as reported by TechCrunch, will provide regularly updated, cleaned, and structured datasets in machine-readable formats (e.g., JSON). This will not only simplify the technical work for AI teams but also solve two key problems: it will improve the quality of training sets and ensure the transparency of data provenance, which is critical for license compliance and combating model "hallucinations." The initiative aims to enhance the overall reliability and responsibility in the AI development ecosystem.
Wikimedia Launches Project to Make Wikipedia Data More Accessible to AI
