The Wikimedia Foundation will have large technology companies pay for the use of Wikipedia content when training AI models.
Wikimedia announced licensing agreements on Thursday with Microsoft, Meta, Amazon, Perplexity, and Mistral AI. This means these companies are joining the commercial Wikimedia Enterprise program, which offers API access to more than 65 million Wikipedia articles at higher speeds and volumes than free public APIs.
‘Wikipedia crucial for AI development’
Google already agreed to a similar agreement in 2022, writes Ars Technica. Financial details have not been disclosed. According to Wikimedia Enterprise President Lane Becker, Wikipedia is a crucial data source for AI development, but large-scale use also entails significant infrastructure costs. The additional revenue should help compensate for those costs.
Wikipedia founder Jimmy Wales is positive about AI training with Wikipedia data but emphasizes that companies must also contribute financially. “Free knowledge is not free to maintain,” he stated.
Increasing pressure from AI scraping
In April 2025, Wikimedia reported that bandwidth consumption for multimedia increased by 50 percent, with bots responsible for the majority of the heaviest infrastructure requests. At the same time, human traffic decreased by about eight percent, which is disastrous for the donations it needs to survive.
read also