Anthropic's Claude Model Requested Code for Paid Dashboard in Georgia

According to Axios, Anthropic's AI model Claude requested code to query a dashboard containing information normally accessible for a fee in the US state of Georgia. Authorities reported that the model did not access the voter registration system and did not make any changes to voting records.

An Anthropic spokesperson stated that Claude was used to collect public data as part of an internal statistics project. The spokesperson emphasized that although the model was not asked to do so, it requested code to access information that is normally paid for, but it did not access the state's voter registration system, nor did it modify or scrape any voting records. It was also stated that necessary measures have been taken to prevent such actions.

A source with knowledge of the matter speaking to Axios claimed that Anthropic had informed the Georgia Secretary of State's Office and the White House about the situation weeks ago. The same source claimed that US administration officials requested an earlier statement from the Secretary of State regarding the incident, but these calls were ignored. This incident has brought back concerns regarding the capabilities of artificial intelligence systems and the extent to which they can be controlled by humans.