Artificial intelligence bot accesses Georgia voting data

An artificial intelligence agent accessed publicly available Georgia voter data, sources from the Secretary of State’s Office and AI company Anthropic confirmed.

On Monday, Anthropic confirmed its Claude artificial intelligence bot accessed the data without being prompted to do so. A company spokesperson said Claude did not access the state’s voter registration system, change voting records or obtain any individual voter records.

“The model pulled only aggregate counts, which can be derived from voter data the state makes available to anyone for a fee,” the spokesperson said.

The Secretary of State’s Office confirmed the AI model did not enter any voter files or Georgia’s voter registration system. The election data it’s said to have accessed is publicly available via the state’s data hub, officials said.

“Anthropic’s AI agent accessed publicly available voter information, just as anyone else in this country can,” said Robert Sinners, a spokesperson for the office.

“There was no data breach, hack or unlawful or harmful tampering with voter information or election systems and zero reason to make any changes to our public facing website,” he said.

The disclosure comes amid rising concerns about artificial intelligence agents gaining access to information without being asked to do so. A new Atlanta Journal-Constitution survey found a majority of Democrats and Republicans support greater regulation of artificial intelligence.

Anthropic officials appeared to reference the Georgia incident in an Oct. 9 blog post detailing various incidents it deemed to have “minimal real-world impact.” It said a researcher had used Claude Mythos 5 on a statistics project that required it to pull public data from an unidentified agency’s web site.

Officials have not said when the Claude incursion happened. The Claude Mythos 5 large language model debuted in June.