In the era of generative AI’s frantic absorption of open data, Wikipedia has officially transformed the relationship of “being freely taken” into a commercial agreement: from Microsoft, Google, Amazon to emerging AI companies, all now must access Wikipedia through content licensing and paid services.
(Background: Will Grok replace Wikipedia? Elon Musk reveals xAI is developing “Grokipedia”: a significant improvement over Wikipedia)
(Additional context: V神’s first review of LLM: Grok essentially saves the X platform by “helping truth dissemination,” but still has many hallucinations)

Table of Contents

• AI giants are no longer just “data scraping”
• Why does Wikipedia have bargaining chips?
• Emphasizing that human editors cannot be replaced

The Wikimedia Foundation celebrates the 25th anniversary of Wikipedia’s founding with a series of activities and technical updates, while also clearly signaling externally that the world’s largest online encyclopedia is not just a “free knowledge base,” but a key infrastructure that has signed content licensing agreements with multiple AI giants and officially entered into commercial negotiations.

Currently, Wikipedia has over 65 million articles, covering more than 300 languages, with nearly 15 billion monthly page views. It is the only platform among the top ten most trafficked websites operated by a non-profit organization, and also one of the most important high-quality open data sets for large language models.

AI giants are no longer just “data scraping”

In recent years, with the rise of generative AI, technology companies’ reliance on Wikipedia content has rapidly increased. To respond to this demand and maintain financial sustainability, Wikimedia developed the commercial product Wikimedia Enterprise, which specializes in providing large-scale content reuse and distribution services.

In its latest statement, the foundation disclosed that companies such as Ecosia, Microsoft, Mistral AI, Perplexity, Pleias, ProRata, among others, have become new partners, joining the ranks of existing tech giants like Amazon, Google, and Meta.

This means that companies accustomed to directly scraping Wikipedia content for search results or AI training are now beginning to access data through “licensing cooperation.” Wikimedia Enterprise provides APIs or data streams based on the company’s latency, stability, and data format needs, and companies pay fees to Wikimedia Foundation to support non-profit operations and infrastructure investments.

Why does Wikipedia have bargaining chips?

The Wikimedia Foundation emphasizes in its announcement that Wikipedia has been evaluated as one of the “highest quality” open data sets for large language model training. This is because its content is maintained by approximately 250,000 active volunteer editors, adhering to strict standards such as neutrality, verifiability, and reliable sources, and has undergone long-term version history and community review. These are structural assets that model developers find difficult to reconstruct on their own.

For AI companies, obtaining Wikipedia content is not only about licensing legality and ethical considerations but also about the quality of model outputs and the ability to grasp facts; for Wikimedia, it is about transforming the originally passive traffic scraping into a predictable revenue source to sustain long-term investments in servers, multilingual communities, and technological development.

Emphasizing that human editors cannot be replaced

Interestingly, although Wikimedia has reached content licensing agreements with multiple AI giants, it repeatedly emphasizes in its AI strategy that “humans come first.” The role of AI is to assist volunteer editors, not to replace them.

The foundation plans to use AI to detect destructive edits, flag potentially problematic articles, assist in translation and content discovery, allowing editors to focus on source interpretation, writing, and community governance.

CEO Maryana Iskander states that Wikipedia’s core value lies in “human-driven” knowledge production. Even in the AI era, the platform will maintain a governance structure led by a global volunteer community, with AI tools serving only as aids to lower participation barriers rather than taking over content decision-making.