Wikipedia seeks AI licensing deal to cover rising costs

Wikipedia is seeking more licensing agreements with major technology firms to offset rising costs caused by AI companies using its content, co-founder Jimmy Wales said.

The nonprofit Wikimedia Foundation, which operates Wikipedia, signed a deal with Google in 2022 to allow paid access for training AI models and is in talks with other companies.

Wales said automated scraping by AI bots has increased Wikipedia’s server and memory costs, which are not covered by the small public donations the foundation relies on.

While Wikipedia’s content remains free for individual use, for-profit firms training AI models create a disproportionate financial burden.

Wales noted Wikipedia may consider technical tools to limit AI crawling, though this is challenging due to its commitment to open access, and he emphasized the potential impact of public pressure.

🔗 Source: Reuters

🧠 Food for thought

Implications, context, and why it matters.

Wikimedia Enterprise revenue is modest relative to overall funding; AI-related cost pressures are not quantified

Wikipedia signed a licensing deal with Google in 2022 and is pursuing others. The Wikimedia Foundation’s 2023-2024 budget says Wikimedia Enterprise, the paid Application Programming Interface (API) and services initiative, generated $3.2 million in annual recurring revenue as of January 2023 1. That equals 1.7% of the $185.3 million in support and revenue the Foundation recorded in 2023 2.
Technology makes up 45% of the operating budget for hosting and infrastructure 3. The Foundation ended FY2024 with $271.5 million in net assets 2, yet filings do not break out server-cost jump tied to AI scraping 3 4. The impact of AI crawling on sustainability remains unclear given flat budget growth 1.

AI developers face unsettled obligations when using CC BY-SA licensed content for training

Creative Commons Attribution-ShareAlike (CC BY-SA) licenses require attribution, and they compel sharing adaptations under the same license 5. Some regions allow AI training under copyright exceptions 6. Companies that monetize outputs built on Creative Commons licensed content face murky rules on attribution and whether ShareAlike, the requirement to license derivatives on the same terms, applies 5 7.
This opens a lane for attribution-tracking systems and provenance tooling 5. These tools record data sources and usage history to document compliance during training and inference, when a model generates outputs. Investors can track companies building automated licensing layers that guide developers through ShareAlike duties when model outputs include CC-licensed content 7.