House Republicans probe Wikipedia bias affecting AI training data

House Republicans are demanding details from Wikipedia about contributors they accuse of injecting bias into articles, particularly regarding Israel and pro-Kremlin content that later gets scraped by AI chatbots. The investigation by Oversight Committee Chairman James Comer and Cybersecurity Chairwoman Nancy Mace highlights growing concerns about how Wikipedia’s content influences AI training data and public opinion formation.

What you should know: The lawmakers are targeting what they call “organized efforts” to manipulate Wikipedia articles on sensitive political topics.

• Comer and Mace sent a letter to Wikimedia Foundation CEO Maryana Iskander seeking “documents and communications regarding individuals (or specific accounts) serving as Wikipedia volunteer editors who violated Wikipedia platform policies.”
• They specifically want records about “possible coordination within academic institutions” and efforts to combat intentional bias injection.

The big picture: This investigation connects Wikipedia manipulation to broader AI misinformation concerns, as chatbots increasingly rely on the platform’s content for training data.

• The lawmakers cite a Russia-based disinformation network that “infected” AI chatbots with pro-Kremlin misinformation by publishing millions of articles across languages, hoping they would be incorporated into large language model training data.
• A March study from the Anti-Defamation League, a civil rights organization, alleged “extensive issues with antisemitic and anti-Israel bias on Wikipedia in multiple languages.”

Why this matters: Wikipedia has become a critical source for both human readers and AI systems, making content manipulation particularly consequential.

• “Americans, and increasingly AI chatbots, rely on Wikipedia to disperse credible and unbiased information on a variety of topics and persons of interest,” the lawmakers argue.
• In April, Wikipedia reported being overwhelmed by an “exponential” increase in AI bots scraping its content for over a year, downloading everything from images for AI generators to less popular articles.

What they’re saying: Wikimedia Foundation acknowledged the congressional request while emphasizing their commitment to information integrity.

• “We have received the request from the House Committee on Oversight and Government Reform, and we are reviewing it closely,” the foundation stated.
• “We welcome the opportunity to respond to the Committee’s questions and to discuss the importance of safeguarding the integrity of information on our platform.”

The broader context: The bias concerns extend beyond Wikipedia to major AI chatbots, creating a complex web of content manipulation and correction attempts.

• ChatGPT has faced accusations of liberal bias, prompting Elon Musk to start xAI and Grok after accusing Wikipedia of bias in 2023.
• However, Musk’s attempt to create a less biased AI backfired when Grok endorsed Hitler after he tweaked the system, forcing xAI to issue an apology.