Jimmy Wales, co-founder of Wikipedia, recently clearly stated that current artificial intelligence is not yet trustworthy enough to directly participate in editing entries on the platform. He openly admitted that although the new generation of AI models have made progress in reducing "hallucinations," the phenomenon of AI confidently outputting false information remains "very, very serious." Therefore, Wikipedia will not allow AI to directly participate in the editing process.

Wikipedia

AI can be a "clock," but not an "editor"

Wales added that AI agents are not entirely useless. It can serve as a reminder, helping the Wikipedia community, composed of millions of human editors, to pay attention to small-scale news that might otherwise be overlooked. However, in core content editing, humans remain the only trusted party. Wales emphasized, "We won't let it directly participate in editing because you really can't fully trust it."

Human traffic drops 8%, but Wikipedia does not rely on traffic for survival

At the same time, an undeniable phenomenon is taking place: major AI platforms are relying on Wikipedia content to answer user questions, leading to an overall increase in visits from AI crawlers, while human user traffic has dropped by 8%. Wales, currently a member of the board of directors of the Wikimedia Foundation, believes this change is "worth noting" but "not a disaster."

The reason is that Wikipedia was created in 2001 and mainly relies on user donations to sustain operations, and its business model does not directly depend on traffic. However, Wales has called on AI companies to "take their fair share of costs," as sending millions of requests to Wikipedia incurs real server expenses. He said that Wikipedia has achieved "great success" in signing cooperation agreements with several tech giants and has started blocking those platforms that don't follow the rules, but the subsequent effects remain to be seen.