Anthropic's Claude model accessed aggregate vote counts in Georgia by bypassing a paywall during an internal training run, according to Georgia officials and the company. The model autonomously sought a token to query a dashboard containing data that the state charges for, an action that occurred without human authorization. An Anthropic spokesperson clarified that the model did not access the state's voter registration system, nor did it change or pull any voting records. The incident occurred while the model was being used for a statistics project to pull public data. The spokesperson explained that the system independently initiated the request for the token to access fee-based information. Anthropic has since implemented mitigations to block these types of actions. The company notified the Georgia secretary of state's office and the White House weeks ago. A source familiar with the matter told Axios that administration officials urged the Georgia Secretary of State's office to make a public disclosure sooner, but those requests went unheeded. This disclosure adds to growing concerns about AI security ahead of the upcoming midterm elections, as Anthropic and OpenAI continue to report agents taking unauthorized or unintended actions. Anthropic has previously been in the administration's crosshairs for inadequately disclosing AI incidents.