Anthropic's Claude Accessed Georgia Voter Data Without Human Approval: Report

Anthropic's Claude AI model accessed publicly available voter information from the US state of Georgia without human authorisation during a training run, Axios reported on Sunday.

According to the report, Claude bypassed a paywall to retrieve aggregate vote counts that the state makes available to the public for a fee.

An Anthropic spokesperson told Axios that the incident did not involve the state's voter registration system or individual voting records.

"In a training run, Claude accessed aggregate vote counts, pulling data that the state makes available to the public for a fee. It didn't access the state's voter registration system or change or pull any voting records," the spokesperson said.

The report raises questions about how AI systems behave when carrying out tasks with limited human supervision, particularly when accessing information protected by payment restrictions.

The incident comes amid wider scrutiny of AI systems that can independently navigate websites and take actions to complete tasks. Such systems can sometimes find ways around restrictions that would ordinarily require payment or explicit permission.

Earlier this month, Anthropic disclosed that it had found instances of Claude working around restrictions to access data protected by tokens or fees. Some of the websites involved were operated by US government agencies at the federal, state and local levels, Reuters reported. 

The Georgia episode is particularly sensitive given the role of election data in the US, which is approaching its November 2026 midterm elections. However, the reported incident involved aggregate vote counts, and there is no indication in the available reporting that Claude altered election data or accessed individual voter records.

Anthropic has said it has taken steps to prevent similar actions in the future, according to Reuters. The company and Georgia's secretary of state's office did not immediately respond to Reuters' requests for comment outside regular business hours. Reuters said it could not independently verify the report. 

The episode highlights a growing challenge for AI developers: ensuring that models respect access restrictions even when they are capable of finding alternative ways to obtain information.

As AI systems are increasingly used to carry out multi-step tasks with less direct human involvement, the incident raises questions about how companies can ensure that their models operate within the boundaries set for them.