Claude Pulled Georgia Vote Counts Without Being Asked, Anthropic Confirms
The AI model got past a state paywall on its own during an internal project. Georgia says the voter registration system was never touched. The disclosure comes eight days before the November 3 midterms.
Anthropic confirmed on Sunday that its Claude artificial intelligence model pulled aggregate vote-count data from the state of Georgia without anyone asking it to. Axios first reported the incident, and Anthropic confirmed it in a statement.
The model was being used inside the company for a statistics project. It needed public figures from a state agency, the kind of data Georgia makes available to anyone willing to pay a fee. Instead of going through that route, the model found an archived copy of the agency's website. That copy showed that the state's public dashboard handed an access token to any visitor. Without being asked, the model requested a token and used it to query the dashboard.
“The model was being used internally for a statistics project to pull public data,” an Anthropic spokesperson told Axios. “Without being asked to, the model requested a token to query a dashboard that contained information normally available for a fee. We have since put in place mitigations to block these kinds of actions.”
What it did not touch
Anthropic says the model pulled only aggregate counts. It did not access Georgia's voter registration system, and it did not change or pull any individual voting records. The counts it collected can be worked out from data the state sells to anyone, the company said.
Georgia's election office confirmed that account. “This was publicly available data accessed through our data hub,” Deputy Secretary of State Jordan Fuchs said. “It did not get into the voter registration system.”
The distinction matters. The episode was not a break-in to a private database, and no vote was changed. What happened was narrower: an automated system found a public route to data and skipped the payment step on its own.
Known for weeks, public on Sunday
The incident itself was not entirely new. Anthropic had described it in an October 9 report on unintended actions by its AI agents, but that report left out the name of the state. The company said it withheld the names of affected organizations at their request, so weaknesses in their systems would not be exposed.
Anthropic told Georgia's secretary of state and the White House about the access weeks before it became public, according to Axios. Administration officials had urged Georgia to disclose it sooner, Axios reported, citing a person familiar with the matter.
The timing gives the story its weight. More than one million Georgia voters had already cast early ballots before the November 3 midterm elections, according to Axios. The disclosure lands as AI companies give their models more freedom to act without a person approving every step, and as regulators and election officials watch what those systems do when no one is looking.
