Anthropic says Claude.ai will soon answer US voting questions with state-specific information from Democracy Works, instead of only pointing to a link. Its Oct. 14 midterms post says the change starts later this month for US users, alongside live election results once polls close.
What Claude.ai is adding
Anthropic says Claude will start drawing official voting information from Democracy Works into its Claude.ai responses for US users. A question about where to vote or what ID to bring should get an answer specific to the person's state, without leaving the conversation. It says the data comes from public state and local election sources.
Anthropic says Claude.ai will keep showing a banner linking to TurboVote, a nonpartisan Democracy Works resource, when people ask about voter registration, polling places or election dates. It also says Claude will show up-to-date official results on election night and the days after, when people ask for vote counts.
What Anthropic says it sees in US conversations
The figures here are Anthropic's own analysis of about 250,000 US Claude.ai consumer conversations from Sept. 21 to 27, which it says Claude classified using a privacy-preserving method. Anthropic says about 5.75% touched politics, policy or social issues, and about 0.51% were specifically about elections or voting.
Anthropic says requests asking whom or what to vote for appeared in about 0.01% of those conversations, and requests that broke its election integrity policies in about 0.002%. It says it expects the share of election queries to rise nearer Election Day. We cannot check any of these numbers independently.
Bias tests and misuse
Anthropic says Claude Opus 5.5, Sonnet 5.5 and Haiku 5.5 score 99% on its own political evenhandedness evaluation. It also says outside reviewers found Claude sometimes added unsolicited caveats and counterarguments, and that one found an older model, Sonnet 4.5, softened arguments for right-leaning positions more often. Anthropic says it is addressing those findings.
Anthropic says researchers at the Institute for Strategic Dialogue and the McCain Institute found Claude reliably refused a single harmful request but could miss a malicious campaign split into many harmless-looking tasks. It says it has added classifiers for multi-step misuse and banned accounts tied to a foreign-run fake news site.
How OpenAI and Google compare
We found no 2026 midterm post from either company. OpenAI's 2024 elections post says ChatGPT directed people asking where or how to vote to CanIVote.org, and pointed those asking for results to news sources such as the Associated Press and Reuters. The cited 2024 post does not establish OpenAI's approach for 2026.
OpenAI's usage policies list political campaigning and election interference among uses it does not allow. Google's generative AI policy bars misleading claims about democratic processes made to deceive. Neither page describes in-chat state voting data like Anthropic's plan.
Where to check yourself
Anthropic says the feature will start later this month, so for now registration and rules come from official sources. Vote.gov, a US government site, lets you register to vote in your state. Check your state or local election office for polling places and ID rules before Election Day.
For another AI-misuse story, see researchers' report on an AI agent fleet hitting a map service.
