Wikipedia Founder: AI Can Generate Answers, But Trust Remains Human

Key takeaway

• Jimmy Wales, founder of Wikipedia, argues that artificial intelligence can generate plausible-sounding answers but cannot yet be trusted to create reliable knowledge because of hallucinations—where AI invents false details that sound credible.

• While Wikipedia's content is freely available for AI training, the platform maintains strict sourcing standards and human verification as its core defense against misinformation, remaining sustainable through millions of small donations rather than corporate dependence.

3 Key Points

FAQ

Why doesn't Wikipedia charge AI companies to use its content?

Under Wikipedia's open-source-style licensing model, anyone can use, modify, and redistribute its content for free—commercial or non-commercial. Wales notes that the model couldn't charge even if it wanted to, but he believes it is right for AI to be trained on Wikipedia data, as long as companies recognize their responsibility and manage the infrastructure burden fairly.

What specific problems does Wikipedia face with AI-generated content?

AI hallucinations produce errors that sound convincing rather than obviously false—for example, inventing ISBN numbers, fabricating book titles, or getting relatives' names wrong. One German Wikipedia editor unknowingly used an AI tool to generate book references, resulting in citations to books that did not exist. Because the errors are plausible, they are harder to catch than absurd claims.

How many people actively maintain Wikipedia?

Among regular editors, there are roughly 60,000 to 80,000 globally, with a smaller group of around 5,000 highly active users doing most of the work.