Four Democratic members of Congress have written to the heads of 25 technology companies, including Alphabet, Meta, Microsoft, OpenAI and TikTok, warning that some AI chatbots give wrong answers to election questions too often and that platforms are pulling back from commitments to protect voters, Nextgov/FCW reported Tuesday. The letter, sent two weeks before the Nov. 3 midterms, signals that lawmakers increasingly see accurate voting information as a safety obligation for AI providers, not only for social media platforms.
What Happened
The letter was signed by Sen. Mark Warner of Virginia, the top Democrat on the Senate Intelligence Committee; Sen. Alex Padilla of California; Rep. Jim Himes of Connecticut, the top Democrat on the House Intelligence Committee; and Rep. Joe Morelle of New York. Padilla serves on the Senate Rules Committee and Morelle on the House Administration Committee, the panels that oversee federal elections.
According to Nextgov/FCW, the lawmakers cited "concerning signs of backsliding on some of your commitments," and wrote that some of the companies' chatbots have "an unacceptably high error rate" on election and voting questions, which they said could keep people from voting. The outlet pointed to a September study by the Institute for Strategic Dialogue, a research group that tracks disinformation, which found that Meta's Muse Spark model twice gave Nov. 4 as the date of the midterms, a day after the actual election.
The lawmakers asked the companies to:
- Treat deceptive claims about voting procedures, such as registration deadlines, polling places, ID requirements and mail-in ballot rules, separately from protected political speech, and remove "demonstrably false" voting information once they know about it.
- Give election officials dedicated tools to report false information.
- Use independent fact-checkers alongside internal teams and automated systems.
- Notify users who engaged with content later found to be false or misleading.
- Grow, rather than shrink, election-integrity and trust-and-safety teams.
- Give independent researchers access to products and data needed to assess whether safeguards work.
The letter also faulted companies for slow and inconsistent removal of Russian disinformation that breaks their own rules, Nextgov/FCW reported. The outlet did not report any deadline for responses.
Why It Matters
Chatbots now answer voting questions that people once took to search engines or election offices. A wrong date or deadline from an AI assistant carries the company's credibility, and the letter frames such errors as an enforcement and staffing problem rather than a technical quirk. For boards and executives at AI companies, it signals that election accuracy is becoming part of how lawmakers judge whether safety commitments are real.
The request for independent researcher access is also significant. It would let outsiders measure chatbot error rates directly, which could feed future oversight.
Context
The warning comes as federal support for election defenders has narrowed. Nextgov/FCW noted that the Trump administration has dismantled several offices focused on foreign influence and election disinformation over the past two years, while agencies say election-security support continues. State officials have described fewer federal links this cycle, and the FBI and the Cybersecurity and Infrastructure Security Agency have warned of AI-polished phishing aimed at election officials.
What to Watch
Nextgov/FCW reported no replies from the companies; it sought comment from trade groups TechNet and the Computer & Communications Industry Association. Watch for company responses, any changes to how chatbots handle voting questions before Nov. 3, and whether lawmakers follow with hearings or legislation after the election.
