Wikipedia becomes fee-based for AI bots
The Wikimedia Foundation wants to secure Wikipedia's survival in the AI era. Those who profit should pay for the effort they cause.
(Image: Gerd Altmann, gemeinfrei (Creative Commons CC0))
Wikipedia and related projects of the Wikimedia Foundation are a goldmine for AI companies. They constantly scrape the valuable data to feed their large language models. This causes considerable costs for Wikimedia, which is why they wanted to slow down AI bots. Since the bots increasingly disguise themselves as humans, this has only worked to a limited extent. Now Wikimedia is trying a separate interface (API).
It will be marketed as Wikimedia Enterprise and allows automated, structured querying of content. Occasional use is free of charge, while intensive use incurs fees. These revenues are intended to help finance server operations and ongoing legal defense.
The exact tariffs are apparently a matter of negotiation and depend on whether the customer wants to receive all updates in real-time, in hourly or daily compilations, or only on active request. Commitments regarding availability and speed of response to support requests are also relevant factors.
Money alone is not enough. Customers must, if they use the data compiled and structured data by countless volunteers, cite the source. Royalties for the content itself are not charged.
Advertising
“In a world increasingly drowning in AI, Wikipedia's human knowledge is more valuable to the world than ever,” states Wikimedia's presentation of the Enterprise offering. It emphasizes that AI would die without human-provided information; over time, predominantly AI-generated content would have to be used for AI training, leading to the collapse of AI models.
Videos by heise
Humans contribute to the creation of knowledge in a way that cannot be replaced by AI. While current AI tools can write summaries, they are incapable of discussion and consensus-building, as volunteer Wikipedia editors do daily. Artificial intelligence “is unable to unearth something buried in an archive, nor can it take a photo of an event or an unrecorded place to improve knowledge.”
Moreover, Wikipedia is online in over 300 languages, often written by native speakers. This multilingualism supports the development of inclusive AI models with cultural competence.
(ds)
