Facing increased demands from AI data consumers and pointed questions from US authorities, the Wikimedia Foundation has outlined its strategy for using artificial intelligence over the next three years, emphasizing that the technology will augment, not automate away, the work of Wikipedia’s human volunteer base.
The plan, detailed by Wikimedia’s Director of Machine Learning Chris Albon and Head of Research Leila Zia, presents AI primarily as a tool to streamline workflows and remove technical obstacles for the global community editing the encyclopedia. This approach aims to preserve the human judgment central to the project while leveraging AI for efficiency.
This internal focus arrives as the Foundation, which oversees Wikipedia and related projects with assets over $255 million and more than 700 staff/contractors, navigates significant external forces. The voracious appetite of AI models for training data has heavily taxed Wikipedia’s servers.
Reports in mid-April suggested automated scraping drove bandwidth usage up nearly 50% in the past year, with other figures indicating bots account for up to 65% of resource-heavy traffic. To offer a more structured alternative, Wikimedia Enterprise released datasets on Google’s Kaggle platform on April 16.
Kaggle is a popular online community for data scientists. This initiative provides pre-parsed content (around 25GB zipped) in JSON format, including fields like article titles, IDs, version details, Wikidata QIDs, abstracts, descriptions, image links, infoboxes, and sections (excluding other media, lists, tables, references).
The data is offered under open licenses like Creative Commons Attribution-Share-Alike 4.0 and the GNU Free Documentation License for AI developers, including smaller outfits and individual researchers, potentially easing server load. This builds on prior Wikimedia Enterprise deals, such as those with Google and the Internet Archive established in June 2022.
Navigating Political Headwinds
Just days ago the Foundation received a letter from Ed Martin, the acting US Attorney for DC, questioning its 501(c)(3) nonprofit status. As reported previously, the inquiry demanded responses by May 15, 2025, to a dozen questions about editorial practices and safeguards against foreign influence.
Martin alleged that “Wikipedia is permitting information manipulation on its platform, including the rewriting of key, historical events and biographical information of current and previous American leaders, as well as other matters implicating the national security and the interests of the United States.”
His letter further claimed his office had information suggesting Wikipedia’s “policies benefit foreign powers” and that the board being “composed primarily of foreign nationals” subverts taxpayer interests. He argued that “Masking propaganda that influences public opinion under the guise of providing informational material is antithetical to Wikimedia’s ‘educational’ mission.” Sources close to Martin indicated concerns were particularly focused on edits related to the Israel-Hamas conflict.
This official scrutiny aligns with wider criticism from figures like Elon Musk (who posted “Stop donating to Wokepedia”) and Wikipedia co-founder Larry Sanger regarding perceived bias.
Think tanks like the Manhattan Institute and activist groups such as the Media Research Center have published analyses claiming evidence of bias.
The Anti-Defamation League, designated by Wikipedia editors in 2024 as generally unreliable on the Israeli-Palestinian conflict, issued a report in March 2025 alleging biased editing campaigns by “at least 30 editors” and urging platforms to “deprioritize unvetted Wikipedia’s content on issues related to Jews, Israel and the Middle East conflict.”
Responding to Martin’s letter on April 25, Wikimedia reiterated its goal is to “inform, not persuade,” relying on volunteers and neutrality policies. Critics like editor Molly White characterized the inquiry as “weaponizing laws to try to silence high-quality independent information.”
Wikipedia co-founder Jimmy Wales, addressing bias concerns in a December 2024 interview noted, “It is something I look at, focus on, and think about… But whenever I try to find problematic examples, it’s pretty hard,” referencing the site’s Neutral Point of View policy.
AI as Assistant, Not Replacement
Against this backdrop, Wikimedia’s AI strategy strongly reaffirms the centrality of human judgment. “The community of volunteers behind Wikipedia is the most important and unique element of Wikipedia’s success,” the Foundation’s announcement states, adding that AI cannot replace their “care and commitment to reliable encyclopedic knowledge.” The plan focuses on using AI where it excels: automating tedious tasks, aiding information discovery, facilitating translation, and helping new user onboarding. Specifically, AI tools will assist moderators and patrollers with workflows supporting knowledge integrity; help editors find information faster, leaving more time for deliberation; automate translation of common topics to broaden perspectives; and scale mentorship for newcomers.
Guiding Principles for AI Integration
The Foundation outlined core principles for its AI work, detailed further on Meta-Wiki. The approach will be human-centered, prioritizing human agency. There’s a preference for using open-source or open-weight AI models, aligning with the movement’s ethos.
Transparency in how AI is used is another key tenet. Furthermore, the strategy calls for a nuanced approach to multilinguality and adherence to existing Wikimedia values and policies, including those concerning privacy and human rights, as well as its general Terms of Use. Director of Machine Learning Chris Albon emphasized this framework in, stating: “We believe that our future work with AI will be successful not only because of what we do, but how we do it.”
The claims of political bias and blatant disinformation are true. I find Wikipedia no longer reliable. I can edit on the site, but the amount of editing that needs to be done is overwhelming.