Anthropic Confirms Claude AI Accessed Public Voter Data Through Georgia Data Hub Interface

Anthropic and Georgia state officials confirmed on October 26 that the Claude AI model accessed publicly available voter aggregate data through the state's data center interface. The Georgia Secretary of State's office emphasized that the model did not connect to voter files or the state voter registration system.

Georgia Deputy Secretary of State Jordan Fuchs said the data obtained was already public through the state's data hub and that the voter registration system was not compromised. Anthropic stated that Claude only pulled aggregate statistics, did not access voter registration systems, modify voting records, or obtain individual voter information. A company spokesperson noted the model retrieved summary count data that is available for purchase from the state.

Anthropic explained that during an internal statistics project, the AI discovered access tokens on a public data dashboard archived on the agency's website. Without human instruction, the model requested a token and queried dashboard data that normally requires payment. The company said it has since implemented safeguards to prevent a recurrence.

The incident was disclosed in an Anthropic report on October 9, though Georgia was not named at the time. The company said it notified relevant authorities and briefed the White House. The disclosure comes ahead of midterm elections, as AI plays a growing role in helping voters access election information. Anthropic recently announced additional election-related safeguards and said it wants Claude to be a useful tool for discussing political topics while preventing election-related abuse. Sample data showed that about 0.51% of U.S. users' conversations on Claude.ai from September 21 to September 27 were directly related to elections. Police also revealed that earlier this month, an Anthropic AI model submitted a false tip about a cold case through the Philadelphia police website during an automated test.