Science & Tech · AI · 2 hrs ago

AI chatbots give voters misleading information about midterms, researchers find

AI chatbots give voters misleading information about midterms, researchers find

Researchers found errors in popular chatbots’ election answers as US voters seek information ahead of the midterm elections.

An ongoing study tested seven major chatbots with questions about candidates, polling places and voting rules.

In October, 13.9% of answers about voting logistics were outdated or inaccurate, and some errors could cause voters to miss deadlines or required steps.

The rate of errors varied by chatbot, with Claude and Copilot responsible for most of the clear factual errors in the study.

Chatbots may shape views of candidates and issues even when they do not directly tell people how to vote.

Experts say voters should check chatbot answers against reliable official information.

Microsoft said it is refining Copilot’s election responses, while Anthropic said some safeguards may not have been active in the researchers’ testing.

4 sources

Related

What people are saying

Tim ChambersPrimary researcher at Innovating for the Public Good
We've been increasingly seeing the role of AI delivering answers to users in a very authoritative, very convincing way, but with meaningful errors as a point of concern
via NPR
Users of these chatbot systems could be given errors in how they vote or miss a key voting deadline and elements in the whole voting process that could be derailed if they trusted these answers.
via NPR
Michael AcimanAnthropic spokesperson
because IFPG conducted part of its research using an Application Programming Interface (API), some of Claude's key safeguards may not have been active.
via NPR
Abdiaziz AhmedTech policy strategist at the Brennan Center for Justice
It’s best to treat artificial intelligence like a smart friend who’s sometimes right.
via El País English