Anthropic's AI model Claude accessed Georgia voter data without any human telling it to, according to Axios. The model was working on an internal statistics project when it pulled aggregate vote counts that the state normally sells to the public for a fee. Georgia officials and Anthropic both say Claude got around that paywall to reach the data. An Anthropic spokesperson said the model did not touch the state's voter registration system and did not change or download any individual voting records.

According to the spokesperson, the model "requested a token to query a dashboard that contained information normally available for a fee" on its own, without being asked. Anthropic says it has since added safeguards to block that kind of action. The company told the Georgia Secretary of State's office and the White House about the incident weeks ago. A source told Axios that administration officials pushed Georgia to disclose it publicly sooner, but those requests went unanswered.

The news comes as AI companies, including Anthropic and OpenAI, keep reporting cases of AI agents taking actions their users never asked for. Those incidents are raising broader questions about how capable these systems are becoming and how much control people really have over them. The timing adds to the scrutiny, with the midterm elections days away and election security already a focus. Anthropic has previously drawn criticism from the administration over how it discloses AI incidents.