Anthropic has set out how it is protecting its Claude chatbot ahead of the US midterm elections. In a blog post on 14 October 2026, the company shared its own figures on how people use Claude for politics, described new voting-information features, and said it had stopped foreign and money-making efforts to misuse Claude around the election. All the figures below are Anthropic’s own.
How people use Claude for politics
Anthropic says it analysed 250,000 Claude.ai conversations from 21 to 27 September 2026, using what it calls “a privacy-preserving method”. By its count:
- Politics, public policy or social issues make up “about 5.75% of all conversations in the United States on Claude.ai”.
- “About 0.51% of all conversations are specifically about elections or voting.”
- Asking Claude who or what to vote for showed up in “about 0.01% of US Claude conversations”.
- Requests that broke its election rules appeared in “about 0.002% of conversations”.
Voting information
Claude.ai will keep showing a banner linking to TurboVote, a nonpartisan resource from the group Democracy Works, when people ask about registering, polling places or election dates. Anthropic says that “later this month, Claude will start drawing official voting information from Democracy Works into its responses on Claude.ai for users in the US.” That feature is not live yet. Anthropic also says Claude will show up-to-date official results on election night and in the days after.
Political balance
“Claude Opus 5.5, Sonnet 5.5 and Haiku 5.5 are our most even-handed to date, scoring 99% on our political even-handedness evaluation,” Anthropic says. That is the company’s own test, not an independent one.
Anthropic also asked three outside groups to review Claude: The Future of Free Speech, the Foundation for American Innovation and the Collective Intelligence Project, which surveyed “more than 2,000 people in the US, UK, and India”. Anthropic says these reviews found problems too. Claude “sometimes added unsolicited caveats and counterarguments, such as inserting an opposing view into an advocacy email.” The Foundation for American Innovation found that Claude Sonnet 4.5 “was more likely to soften arguments for right-leaning positions.” Anthropic says it is working to fix both.
Testing and misuse
Anthropic says it worked with the Institute for Strategic Dialogue and the McCain Institute to test Claude’s defences in the US, Israel and Brazil. Testers found that Claude “reliably refused a single harmful request, but could miss a malicious campaign when it was spread across many individual requests that looked harmless on their own.” Anthropic says it has added new classifiers to catch “novel and multi-step misuse”.
The company also described two cases. In one, “foreign-based operators of a fake news site used Claude to produce US midterm coverage favoring Russian, Iranian, and North Korean narratives.” In others, people used Claude “to produce fabricated news and other content designed to attract American audiences and generate advertising revenue.” Anthropic says it banned the accounts involved. It did not say how many accounts or who ran them.
Anthropic has also published a guide for political campaigns. It bans uses such as creating fake social media profiles that pose as real people, mass-producing comments, and writing for news sites that falsely claim to be independent.
What this means
- Few people ask AI how to vote. By Anthropic’s count, such requests are very rare.
- Official voting info is coming. Claude will soon answer where-to-vote questions using Democracy Works data.
- Bad actors are trying. Anthropic says it caught foreign and money-making election campaigns, but by its own account Claude can miss attacks that are split into harmless-looking steps.
Sources
- Anthropic, “Preparing for the 2026 US midterms”, 14 October 2026. https://www.anthropic.com/news/2026-us-midterms
