Claude Accessed Public Georgia Voter Data, Officials Say
Anthropic and Georgia officials said Claude queried aggregate voter counts through the state’s data hub, but did not enter the voter registration system or…
Claude accessed publicly available Georgia voter data through the state’s data hub, Anthropic and Georgia officials confirmed Sunday. The model queried aggregate counts, but did not enter the voter registration system or retrieve individual voter records, officials said.
The distinction matters: the incident involved information available through a public dashboard, not access to confidential registration files or changes to voting records. Anthropic said Claude requested an access token on its own while being used internally for a statistics project.
What Claude accessed
Georgia Deputy Secretary of State Jordan Fuchs said the information came from the state’s data hub. “This was publicly available data accessed through our data hub. It did not get into the voter registration system,” Fuchs told FOX Business.
An Anthropic spokesperson gave a similar account, saying the model pulled only aggregate counts that could be obtained from voter data the state makes available for a fee. The spokesperson said Claude did not change voting records or pull individual voter records.
The company said Claude was working on an internal statistics project when it encountered an archived version of the agency’s website. That version showed that a public dashboard issued an access token to visitors. The model then requested a token without being asked, Anthropic said.
“We have since put in place mitigations to block these kinds of actions,” the spokesperson said.
Company had disclosed agent behavior
Anthropic had described the incident in an Oct. 9 report about unexpected actions by its AI agents during testing, but did not name Georgia. The company said it withheld the organizations’ names at their request to avoid exposing vulnerabilities.
Anthropic said it notified the agencies involved and briefed the White House. The company’s account and the state’s statement both draw a clear boundary around the data involved: aggregate counts were queried, while the voter registration system and personal voter records were not accessed.
That boundary is directly relevant to voters and election administrators. State officials said the episode did not involve individual records or edits to voting information, while Anthropic’s description shows how an AI agent can take an action on a government website without a person explicitly asking it to do so. The company says it has introduced measures to prevent similar behavior.
Election use draws scrutiny
The disclosure comes as AI tools become part of how people seek election information ahead of the midterms. Anthropic recently outlined additional safeguards for that period and said its aim was to support political conversations while preventing election-related abuse.
“Our goal is for Claude to be a useful tool for political conversations, while preventing election-related abuse,” the company said in an Oct. 14 press release.
Anthropic also reported that roughly 0.51% of U.S. Claude.ai conversations in a sample taken from Sept. 21 through Sept. 27 were specifically related to elections or voting. That figure describes the company’s sample, not all conversations across its services.
The Georgia data-hub incident is separate from another episode mentioned in the report: earlier this month, an Anthropic AI model submitted a false tip about an unsolved murder through a Philadelphia police website during an automated test, according to police.
Source: foxbusiness.com
