Blog
Anthropic adds official voting information to Claude and details its safeguards for the US midterms
On 14 October 2026 Anthropic said Claude.ai will soon give US users state-specific voting information from Democracy Works and show official results after Election Day. It also reported bias test results, weaknesses found by outside reviewers, new misuse classifiers and banned accounts tied to fake news sites.
On 14 October 2026 Anthropic described how it is preparing Claude for the 2026 US midterm elections. It covers new voting features, bias testing and misuse it has found. All figures below are Anthropic's own.
How people use Claude for politics
Anthropic analysed a random sample of about 250,000 US Claude.ai consumer conversations from 21 to 27 September 2026, using what it calls a privacy-preserving method. It reports that:
- About 5.75% of US conversations touched on politics, public policy or social issues.
- About 0.51% were specifically about elections or voting.
- Requests asking Claude whom or what to vote for were rare, at about 0.01%.
- Requests that broke its election integrity rules appeared in about 0.002% of conversations, based on a separate sample of 250,000 conversations.
Voting information in the chat
When people ask Claude.ai about voter registration, polling places or election dates, Claude will keep showing a banner linking to TurboVote, a nonpartisan service from Democracy Works. Anthropic announced two additions:
- Later this month, Claude will start drawing official voting information from Democracy Works into its answers for US users. Someone asking where to vote or what ID to bring should get an answer specific to their state. Democracy Works collects this information from state and local election sources.
- On election night and in the days after, Claude will show the latest official results when people ask for vote counts.
Political bias testing
Anthropic says Claude Opus 5.5, Sonnet 5.5 and Haiku 5.5 each score 99% on its internal political evenhandedness evaluation, and that they decline fewer political requests than earlier models. It also asked three outside groups to review Claude: The Future of Free Speech, the Foundation for American Innovation and the Collective Intelligence Project.
Anthropic reports that most responses were rated strong, but it also lists shortcomings the reviewers found. Claude rarely refused requests but sometimes added unrequested caveats or counterarguments, such as inserting an opposing view into an advocacy email. The Foundation for American Innovation found that Claude Sonnet 4.5 did not always hedge evenly and was more likely to soften arguments for right-leaning positions. Anthropic says it has improved how Claude handles yes-or-no questions and short-answer requests, where answers could be less balanced, and is working on more consistent use of caveats.
Misuse and enforcement
Anthropic's Usage Policy bans using Claude to deceive voters or disrupt elections, for example by spreading false information about candidates or voting, impersonating election officials or trying to suppress turnout. A new guide for campaigns lists allowed uses, such as press statements and fundraising emails, and banned ones, such as fake social media profiles and sites that falsely claim to be independent news.
Anthropic worked with the Institute for Strategic Dialogue and the McCain Institute to test its defences in the US, Israel and Brazil. They found that Claude reliably refused a single harmful request but could miss a malicious campaign split into many harmless-looking tasks, such as audience research, building a news site and drafting content. Anthropic says it has since added classifiers aimed at this kind of multi-step misuse, improved detection of foreign state actors and strengthened cyber protections for election infrastructure and campaigns.
It also described two enforcement cases. Foreign-based operators of a fake news site used Claude to produce midterm coverage favouring Russian, Iranian and North Korean narratives. Other actors used Claude to produce fabricated news aimed at American audiences to earn advertising revenue. Anthropic says it banned the accounts in both cases.
Why it matters
AI assistants are becoming a place where people look up election facts. Directing questions to official data, and publishing the weaknesses outside reviewers found, makes Anthropic's approach easier to check. For businesses and agencies using Claude for political or civic content, the campaign guide is a practical reference. Anthropic says it plans to publish a fuller account of election-season usage in the coming months.