The Wikimedia Foundation announced new licensing deals with major AI firms to charge for high-volume access to Wikipedia content used to train AI models, citing rising infrastructure costs from large-scale scraping.

The Wikimedia Foundation announced licensing deals with Microsoft, Meta, Amazon, Perplexity, and Mistral AI for access to Wikipedia content used to train AI models. The deals expand participation in Wikimedia Enterprise, which sells higher-speed, higher-volume API access to Wikipedia’s 65 million articles compared with free public APIs. The foundation reported rising infrastructure costs and bot-driven bandwidth growth, and it disclosed an approximately 8 percent year-over-year decline in human traffic after improved bot detection. Wikipedia paused a pilot of AI-generated article summaries after volunteer editors objected, while Jimmy Wales said he welcomes AI training on Wikipedia data but said companies should pay their share of the costs.