Anthropic, the AI company behind Claude, announced this week that it will begin pumping “official” U.S. voting information and election results directly into its chatbot later this month, just in time for the high-stakes November midterms.

The company says the data will come from Democracy Works, a Brooklyn-based nonprofit that runs TurboVote.

Anthropic claims this will help users get state-specific answers on polling places, voter ID requirements, registration, and logistics without leaving the chat. Claude will also display “up-to-date” official results on election night and in the days after.

In a blog post titled “Preparing for the 2026 US midterms,” Anthropic framed the move as a responsible effort to make Claude “a useful tool for political conversations, while preventing election-related abuse.”

The company analyzed roughly 250,000 U.S. Claude.ai conversations from late September and reported that politics, public policy, or social issues already make up about 5.75% of chats. Election- or voting-specific talks account for 0.51%, and direct “whom should I vote for” requests are vanishingly rare at 0.01%.

Anthropic has partnered with Democracy Works since 2024. Earlier this year it already displayed a TurboVote banner for registration and polling questions. Now the chatbot will pull the information straight into responses.

Democracy Works presents itself as nonpartisan. Its funding history tells a different story. The group has received millions from left-of-center sources including the New Venture Fund, Democracy Fund, MacArthur Foundation, Rockefeller Brothers Fund, Pew Charitable Trusts, Knight Foundation, and others frequently tied to progressive election and “democracy” initiatives. InfluenceWatch and grant trackers document these patterns clearly.

In other words, one of Silicon Valley’s leading AI platforms is handing the keys for midterm voting guidance to a nonprofit bankrolled by the usual progressive foundation complex. Americans asking Claude where to vote or what ID is required will receive answers filtered through that pipeline.

Anthropic insists its models score highly on internal “political evenhandedness” tests (99% for the latest Claude versions) and that it has strengthened classifiers against coordinated abuse, foreign influence operations, and deceptive campaign content.

It says it has already banned accounts tied to fake news sites pushing adversarial narratives and financially motivated election spam. Independent reviews it commissioned still flagged issues, including uneven hedging that more often softened right-leaning arguments.