LONDON, U.K.: As it turns 25, Wikipedia is formalizing its relationship with the artificial intelligence industry, striking new commercial agreements that allow major tech companies to access its vast archive of human-curated knowledge at scale.

The Wikimedia Foundation, the nonprofit organization that runs Wikipedia, said on January 15 that it has signed deals with AI companies including Amazon, Meta Platforms, Perplexity, Microsoft, and France's Mistral AI.

Once a symbol of the early internet's promise of free and open knowledge, Wikipedia now sits at the center of debates over how generative AI systems are trained and who bears the cost of maintaining the data they rely on. Aggressive scraping of content by AI developers has intensified questions about fairness, as chatbots and search tools increasingly summarize information rather than directing users to original websites.

The Wikimedia Foundation first began signing AI licensing agreements in 2022, when it struck a deal with Google, and expanded those arrangements last year to include smaller companies such as search engine Ecosia.

Under the latest agreements, AI companies will pay to access Wikipedia content "at a volume and speed designed specifically for their needs," the foundation said, without disclosing financial terms. The aim is to monetize heavy traffic generated by AI developers, whose automated systems place a significant strain on Wikipedia's infrastructure.

Despite legal battles elsewhere over copyright and data use, Wikipedia co-founder Jimmy Wales said he supports AI training on the site's content.

"I'm very happy personally that AI models are training on Wikipedia data because it's human-curated," Wales told The Associated Press. "I wouldn't really want to use an AI that's trained only on X, you know, like a very angry AI," he said, referring to X, owned by Elon Musk.

Wales said Wikipedia wants to collaborate with AI developers, not block them, but added that "you should probably chip in and pay for your fair share of the cost that you're putting on us."

Last year, the Wikimedia Foundation reported that human visits to the site fell 8 percent, while automated bot traffic surged, sometimes using tactics to evade detection as they scraped large volumes of data for AI models. The trend reflects broader changes in how people find information online, as AI-powered summaries increasingly replace traditional search results.

Wikipedia remains the ninth-most visited website globally, hosting more than 65 million articles in 300 languages and edited by roughly 250,000 volunteers. Its free access has been central to its success.

"But our infrastructure is not free, right?" said Maryana Iskander, who is stepping down on January 20 and will be replaced by Bernadette Meehan.

Maintaining servers and technical systems that allow both individuals and companies to draw data from Wikipedia carries real costs, she said. Most of the foundation's funding comes from about 8 million donors, primarily individuals.

"They're not donating in order to subsidize these huge AI companies," Wales said. "They're saying, ‘You know what, actually, you can't just smash our website. You have to sort of come in the right way.'"

Beyond licensing, the foundation sees potential for AI tools to assist editors by reducing routine work, such as fixing dead links or improving search. While AI is not yet capable of writing full Wikipedia articles, Wales said it could support editors and enhance users' interactions with the site.

"You can imagine a world where you can ask the Wikipedia search box a question, and it will quote to you from Wikipedia," he said.

As Wikipedia marks a quarter-century online, it is also navigating political criticism, including attacks from figures on the right who accuse the site of bias and scrutiny from U.S. Republican lawmakers over alleged manipulation of its editing process. Wales dismissed AI-generated rivals as limited.

"Large language models aren't good enough to write really quality reference material," he said.