Claude accesses election dashboard; no data altered

Moving on, an AI system developed by Anthropic has accessed a restricted dashboard containing election figures from the U.S. state of Georgia without explicit human authorization. And this incident has raised fresh questions about whether AI bypasses access restrictions when operating with limited human supervision. Anthropic says its cloud model was working on an internal statistics project when it independently sought a token to access Georgia's election dashboard. And the information was aggregate vote counts, which the state makes available through a paid service. The incident did not involve confidential voter records, access to voter registration system, or changes to election data. the company said the Claude was not instructed to bypass the payment requirement. It has since introduced safeguards to prevent similar actions. Anthropic has also described other instances in which Claude worked around restrictions to access information protected by tokens or payment requirements, including on government websites. The disclosure comes as U.S. officials and technology companies face growing pressure to ensure that AI systems operating with limited supervision do not exceed their authority. According to a reporting by Axios, Anthropic informed Georgia's Secretary of State and the White House about the incident weeks before it became public. With the U.S. midterm elections approaching, this episode has renewed concerns about the security of election infrastructure and the need for effective oversight of autonomous AI systems. While there is no evidence that Claude interfered with the election administration, the incident highlights a key challenge ensuring that AI agents can complete tasks independently without crossing access boundaries or taking actions they were not authorized to take.