Anthropic plans voting information and election results in Claude for US midterms

Anthropic announced additional election features and safeguards for Claude on October 14. The company said US users would gain access to official voting information within Claude.ai responses later this month.

The information will come from Democracy Works, which gathers material from state and local election authorities. Claude will provide state-specific voting guidance and display updated results during election night and afterward.

Anthropic examined 250,000 US consumer conversations from September 21 through 27. It said about 0.51% concerned voting or elections, while roughly 5.75% involved politics, public policy or social issues more broadly.

Testing also exposed a weakness: Claude could reject an explicitly harmful request yet miss an abusive campaign divided into seemingly benign tasks. Anthropic said it added detection systems for such activity and banned accounts involved in fabricated election coverage.

The company plans further work on political balance and election-related testing. In its, Anthropic also said it hopes to publish a fuller analysis of election-cycle usage in the coming months.