Democratic Sens. Mark Warner of Virginia and Alex Padilla of California and Reps. Jim Himes of Connecticut and Joe Morelle of New York warned major technology companies in a letter that their platforms are falling short on commitments to protect elections ahead of the Nov. 3 midterms, according to Bloomberg.
The letter went to the chief executives of Anthropic, OpenAI, Alphabet $GOOGL, Meta $META, SpaceX and others, according to Bloomberg. The recipients number 25 companies, including Microsoft $MSFT and TikTok, according to Nextgov.
"Unfortunately, there are concerning signs of backsliding on some of your commitments to protecting elections from misuse of your products," the lawmakers wrote.
They said some firms have been sluggish about taking down deceptive content that Russia has circulated, and that certain platforms and AI tools let users generate bogus or deceptive election material even though company rules forbid it. They also said some AI chatbots have "an unacceptably high error rate" in their replies to users asking how elections and voting work.
Warner and Himes serve as the senior Democrats on the intelligence panels in their respective chambers. Padilla and Morelle occupy equivalent posts on the Senate Rules and House Administration committees, which have jurisdiction over elections.
The letter asks the companies to remove fake content and work with independent researchers. It also asks them to:
Provide election officials with purpose-built tools for flagging inaccurate claims.
Bring in outside fact-checkers to work with in-house staff and automated tools.
Tell users when material they interacted with turns out to be inaccurate or deceptive.
Take down "demonstrably false" information about when to register, where to vote, what ID is required and how mail-in ballots work as soon as they are alerted to it.
The lawmakers also took aim at companies that have trimmed the headcount of their trust and safety teams. "Refusing to grow and expand your election integrity teams — and even shrinking these teams — does not support your stated commitments to countering harmful election content," they wrote.
A study from the Institute for Strategic Dialogue in September cataloged chatbot mistakes that could lead voters astray. In two of its answers, Meta's Muse Spark model listed Nov. 4 as Election Day for the midterms, one day later than the real vote, according to Nextgov.
Anthropic, Alphabet, SpaceX, Meta and OpenAI had not responded to requests for comment.
It is unclear what consequences the companies could face if they ignore the warning. President Donald Trump has played down the need for tough rules on tech companies, particularly for AI, and has said the companies could police themselves.