A bipartisan pair of House lawmakers is pressing the chief executives of seven leading artificial intelligence (AI) companies to explain how they plan to prevent political bias, misinformation, and inaccurate election-related information from influencing voters during the 2026 midterm elections.
In a May 13 letter, Reps. Mike Lawler, R-N.Y., and Josh Gottheimer, D-N.J., said they want to work collaboratively with the companies to address the risks posed by AI systems that are increasingly being used by tens of millions of Americans to research candidates, ballot measures, and voting procedures.
“AI is going to play a defining role in this election cycle, whether the industry is ready for it or not,” the lawmakers wrote.
“The question is whether companies in this space will get ahead of the problem or be forced to address it after the damage is done”, they said, adding, “We stand ready to work with you and your teams to get this right.”
Reps. Lawler and Gottheimer said the accuracy of AI-generated responses has become “critical for election integrity” as large language models increasingly serve as a gateway to political information, displacing traditional search engines and other established sources of news and voter guidance.
The lawmakers emphasized that while American leadership in AI is a national security priority, it also carries significant civic responsibilities.
“When a single AI system answers political questions for tens of millions of users at scale, a small bias in either direction can shape voter perception in ways no campaign, broadcaster, or newspaper ever could,” they wrote.
They cited research from the Institute for Advanced Study and other organizations showing that leading AI chatbots provided inaccurate or incomplete answers to basic voter questions during the 2024 election cycle, including information about polling locations, voter eligibility, and election deadlines.
The letter also raises concerns about the data used to train AI systems, noting that sources such as Reddit, Wikipedia, YouTube, and Yelp often blend verified facts with unverified opinions and can be manipulated by campaigns, super PACs, and advocacy groups seeking to shape public perceptions of candidates.
“A thriving democracy depends on voters having accurate, accessible, and trustworthy information,” the lawmakers wrote.
“As AI tools increasingly serve as the bridge to that information, displacing traditional search engines, the responsibilities to the public become more comparable to those of a broadcaster or a publisher than to those of a software company,” they said.
The congressmen asked the companies to explain what data sources they use to train and refine their models and how they evaluate the political reliability and partisan balance of those sources.
They also requested details about internal human and automated review systems used to identify and correct biased or inaccurate information about federal, state, and local candidates before it reaches users.
The lawmakers further asked whether the companies have conducted internal or third-party audits of how their models respond to questions about candidates for federal office and whether they will release those findings before November 2026.
They also want commitments to label political content with sourcing and confidence indicators, establish procedures for correcting factual errors identified by campaigns or elected officials, and refrain from down-ranking official campaign, congressional, and verified candidate-controlled websites.
The letter requests written responses from the company officials within 30 days, reflecting the lawmakers’ view that the seven-month window before the midterm elections leaves little time to ensure that AI tools provide voters with reliable and unbiased information.
The letter was sent to Sam Altman of OpenAI, Sundar Pichai of Google, Satya Nadella of Microsoft, Dario Amodei of Anthropic, Mark Zuckerberg of Meta Platforms, Aravind Srinivas of Perplexity, and Elon Musk of X Corp.