Preparations are underway for the 2026 US midterm elections.

Anthropic’s Claude election safeguards are being expanded before the 2026 US midterms, adding voting information, political bias tests, misuse classifiers, and threat-intelligence findings. The update matters because more people are turning to Claude.ai for political and election questions, and the company wants those conversations to remain useful while reducing election-related abuse.

Background: Early 2026 Groundwork

Earlier this year, Anthropic shared an update on its election safeguards, describing how it measures and prevents political bias, detects misuse, and directs people to accurate voting information.

In April, the company said that when people ask Claude.ai about voter registration, polling places, or election dates, the app will show a banner linking to TurboVote, a nonpartisan resource from Democracy Works.

Development: September Data and Policy Updates

A September 21 to 27, 2026 sample of 250,000 US Claude.ai consumer conversations, analyzed through Anthropic Insights using a privacy-preserving method, showed that conversations touching broadly on politics, public policy, or social issues accounted for about 5.75% of all US conversations. About 0.51% were specifically about elections or voting.

The sample suggested that people rarely ask Claude whom or what to vote for. Requests asking for a voting recommendation appeared in about 0.01% of US conversations. Requests that violated Anthropic’s election integrity policies appeared in about 0.002% of conversations.

Most political requests were informational. People used Claude to make sense of geopolitical events, research candidates, draft documents, assist with coursework, conduct academic or professional research on public policy, navigate government services, and weigh arguments for and against policies. Anthropic expects the share of political and election-related queries to increase closer to Election Day.

Last week, Anthropic announced an updated Usage Policy. The update clarifies what its election rules cover and what they do not, and adds examples to help users understand where the lines fall. The policy prohibits using Claude to deceive voters or disrupt elections, including spreading false information about candidates or voting, impersonating election officials, or trying to suppress turnout.

Anthropic also published a guide for political campaigns. It describes permitted everyday uses, such as drafting press statements and fundraising emails, preparing for debates, analyzing turnout data, and building voter outreach materials. It also lists prohibited uses, including creating fake social media profiles that pose as real people, mass-producing comments to inflate a campaign’s engagement, and writing content for news sites or pages that falsely claim to be independent.

Latest Progress: October Voting Information and Defenses

According to the October 14, 2026 policy update, Claude will later this month begin drawing official voting information from Democracy Works into Claude.ai responses for US users. Someone asking where to vote or what identification to bring to their polling place can get an answer specific to their state without leaving the conversation. The information comes from publicly available data that Democracy Works collects from state and local election sources.

Claude will also begin showing up-to-date results on election night and in the days that follow, so users asking for current vote counts or election results can see the latest official numbers directly in their conversation.

Before every model launch, Anthropic tests its models for political bias. Claude Opus 5.5, Sonnet 5.5, and Haiku 5.5 scored 99% on its political evenhandedness evaluation. Anthropic said these models engage more readily with political discussions and decline fewer requests, while still aiming for accurate and evenhanded answers that do not favor a particular political viewpoint.

To pressure-test how Claude handles free expression, Anthropic asked three independent groups to review its responses. The Future of Free Speech, an independent think tank at Vanderbilt University, tested Claude on hundreds of requests about contentious topics. The Foundation for American Innovation compared how Claude responded to matched prompts from the left and right, covering mainstream policy debates and more extreme scenarios. The Collective Intelligence Project surveyed more than 2,000 people in the US, UK, and India to develop evaluation principles and criteria for judging model responses.

Most responses were rated as strong. However, the Future of Free Speech and the Collective Intelligence Project found that Claude rarely refused requests but sometimes added unsolicited 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 said it has improved how Claude responds to direct yes-or-no questions and shorter answers, where its responses could be less balanced. It is also using the findings to improve when and how Claude adds caveats or counterarguments, so requests from across the political spectrum are treated more consistently. The company is continuing to improve tests for uneven hedging and expanding its political bias evaluations to better reflect real conversations, including questions about candidates and voting.

For misuse detection, Anthropic deploys automated classifiers trained to identify election-related abuse. To test how those protections hold up, it partnered with the Institute for Strategic Dialogue and the McCain Institute, both independent research organizations, to pressure-test its protections in the US, Israel, and Brazil. Analysts tested whether Claude would assist with deceptive campaigning, misinformation, and other forms of election abuse. They began with direct requests, then broke full campaigns into smaller tasks that could look harmless on their own, such as audience research, building news sites, drafting content, and planning distribution. They found that Claude reliably refused a single harmful request but could miss a malicious campaign when it was spread across many individually harmless-looking requests.

Anthropic said it has added new classifiers to better identify novel and multi-step misuse, including attempts to build fake social media accounts and news sites aimed at political discussions, or to run financially motivated election schemes. It also enhanced its ability to detect election-related misuse by foreign state actors, strengthened cyber protections for election infrastructure and campaigns, and built targeted safeguards for localized risks identified in the US, Israel, and Brazil, all of which have major national elections this year.

Threat intelligence and enforcement teams review flagged activity and take action when a violation is confirmed, including banning accounts. Anthropic also routinely shares examples of unintended model actions or behavior observed during evaluations and internal use of Claude.

Anthropic’s recent threat intelligence report described a broader shift: political, state-aligned, and commercial actors are incorporating AI into nearly every aspect of influence operations, from planning and content creation to building platforms used to disseminate a message. Because early stages of some operations appear on Claude, Anthropic said it often has a chance to stop them before they reach the public.

In one case, foreign-based operators of a fake news site used Claude to produce US midterm coverage favoring Russian, Iranian, and North Korean narratives. Anthropic also observed threat actors seeking to exploit the election for financial gain, using Claude to produce fabricated news and other content designed to attract American audiences and generate advertising revenue. The company banned the accounts involved and improved its ability to detect this type of abuse.

What to Watch Next

Anthropic says more people are turning to Claude for information about this year’s elections than ever before, and it plans to keep investing in making Claude more evenhanded, preventing and addressing misuse, and being transparent in its actions. The company also expects to share a more detailed account of how people used Claude throughout the election cycle, a signal that future safeguards may be shaped by observed query patterns as the 2026 US midterms approach.

Meta description: Anthropic details Claude election safeguards for the 2026 US midterms: bias tests, Democracy Works voting info, misuse classifiers, and threat findings.

Tags: Anthropic, Claude, US midterms, election integrity, AI safety

Featured image: abstract technology style shield and voting-line network, no real people, no company logos, no text.