Wikipedia co-founder Jimmy Wales has confirmed that the online encyclopedia will not permit artificial intelligence to edit articles directly, citing persistent concerns over AI hallucination. Speaking to AFP on Monday at a climate action week event in London, Wales stated that while newer AI models have reduced the frequency of fabricated outputs, the problem remains "very, very bad." He emphasized, "We would not let it edit directly because you can't really trust it enough."
AI Hallucinations Remain a Critical Barrier
AI hallucination refers to the phenomenon where language models generate confident-sounding but entirely false information. Wales noted that despite improvements in recent models, the risk is still too high for Wikipedia, which relies on accuracy and verifiability. Instead, he suggested that AI agents could be useful in alerting the community of millions of editors to niche news that might otherwise go unnoticed, but not in making direct edits.
Rising AI Bot Traffic and Declining Human Visitors
Artificial intelligence platforms frequently draw on Wikipedia's content to answer user queries, leading to a surge in bot traffic. Wales revealed that human traffic has dropped by eight percent, which he described as "meaningful" but "not a disaster" for the site, which ranks among the world's ten most visited websites. Since Wikipedia's business model relies on donations rather than advertising revenue, the decline in human visitors does not directly threaten its finances.
Calls for AI Companies to Pay Fair Share
Wales encouraged AI companies to "pay their fair share" for the server costs incurred by millions of automated requests. He stated, "Hammering us with millions of requests costs real money." Wikipedia has already secured agreements with several major tech firms, and Wales noted that the platform is "starting to block the ones who aren't behaving themselves," though he added, "we'll see how that goes." The site, founded in 2001, continues to rely on user donations to sustain its operations.
