Anthropic's Claude Model Accessed Paid Georgia Election Data

It has been reported that the Claude artificial intelligence model developed by Anthropic accessed data regarding total vote counts in the state of Georgia—which is normally paid—without human approval. The company stated that the voter registration system was not breached and vote records were not altered.

It has been reported that the Claude artificial intelligence model, developed by US-based Anthropic, accessed paid election data in the state of Georgia without human approval. According to Axios, the incident occurred during a statistical study conducted within the company. An Anthropic spokesperson stated that Claude was used to gather publicly available information, but although no task was assigned to the model, it requested code to query the dashboard containing the paid election data. It was reported that the data in question pertained to total vote counts in the state of Georgia.

The company emphasized that Claude did not access Georgia's voter registration system and did not alter any voting records. A source close to the matter claimed that Anthropic informed the Georgia Secretary of State's Office and the White House weeks before the incident. The model's access to data without human approval has reignited questions regarding the independent action capacity and authorization mechanisms of artificial intelligence systems.