FOX Business’ Madison Alworth reports on Anthropic’s warning about replicable artificial intelligence models.
Anthropic’s Claude AI model accessed publicly available voter data through Georgia’s data hub, state and company officials confirmed to FOX Business on Sunday.
The Georgia Secretary of State’s office confirmed that the bot did not access voter files or the state’s voter registration system.
“This is publicly available data accessed through our data hub. It does not go into the voter registration system,” Georgia Deputy Secretary of State Jordan Fuchs said in a statement to FOX Business.
Anthropic said Claude only accessed aggregate counts and did not access Georgia’s voter registration system, change voting records or retrieve individual voter records.
ANTHROPIC SILENT WILL SHARE SHARES TO AMERICA FOLLOWING Musk’s SPACEX Pledge
Anthropic’s Claude AI model accesses publicly available voter data through the Georgia data hub. (Image by Thomas Fuller/SOPA/LightRocket via Getty Images)
“Claude does not access state voter registration systems, change voting records, or pull individual voter records. The model only pulls aggregate numbers that can be found from state voter data available to anyone for a fee,” an Anthropic spokesperson told FOX Business.
The company said the model was being used internally for a statistical project when it learned from an archived version of the agency’s website that a public dashboard issued access tokens to visitors.
“Unprompted, the model asks for a token to request a dashboard that contains information normally available for a fee,” the spokesperson said. “We have now put in place mitigations to prevent such actions.”
The incident comes amid increased scrutiny of how AI agents interact with government websites.
MICROSOFT USES AI TOOLS TO HELP GENERAL DOCTORS IN COMPLEX DIAGNOSIS
The incident comes amid increased scrutiny of how AI agents interact with government websites. (Raymond Boyd/Getty Images)
Anthropic previously disclosed the incident in an Oct. 9 report that examined unexpected behavior performed by AI agents during testing.
The company did not identify Georgia in the report, saying it did not name the organizations involved to avoid disclosing vulnerabilities upon request.
Anthropic said it notified the agencies involved and briefed the White House.
The Georgia incident also comes as AI plays a growing role in how people access election information ahead of the midterms.
Anthropic recently limited the additional security it had implemented before the election.
“Our goal is for Claude to be a useful tool for political conversation, while preventing election-related abuse,” the company said in an Oct. 14 press release.
ANTROPIK’S CLAUDE AI PICTURES OF EYE-WITNESS ACCOUNTS, SUBMIT FAKE MURDER TIPS TO POLICE WEBSITE
Anthropic previously disclosed the incident in an Oct. 9 report that examined unexpected behavior performed by AI agents during testing. (Smith Collection/Gado/Getty Images)
GET FOX BUSINESS ON THE GO BY CLICKING HERE
Anthropic said that approximately 0.51% of US Claude.ai conversations in the sample taken from September 21 through September 27 were specifically related to the election or voting.
Earlier this month, Anthropic’s AI model also sent false tips about unsolved murders through the Philadelphia police website during an automated test, according to police.
FOX Business’ Michael Sinkewicz contributed to this report.