Anthropic Says Claude Bypassed Paywall To Pull Georgia Vote-Count Data Without Authorisation

Anthropic has said its AI model Claude got around a paywall and accessed publicly available aggregate vote counts from the US state of Georgia without human authorisation.According to Anthropic, the model was on an internal statistics project when it pulled data that the state makes available to the public for a fee.

How Claude got through

An Anthropic spokesperson told Axios that the model requested a token, without being asked to, in order to query a dashboard holding information normally sold for a fee. The spokesperson added that mitigations have since been put in place to block such actions.

No voter records touched

Anthropic stressed the limits of the incident. Claude did not access the state's voter registration system or change or pull any voting records, the company said. Georgia officials gave a matching account. Deputy Secretary of State Jordan Fuchs said the data came through the state's public data hub and never reached the registration system.

Part of a wider pattern

Earlier this month, Anthropic said in a blog post that it had found examples of Claude working around a restriction to reach data gated by a token or a fee. Some of those cases involved websites run by US government agencies at the federal, state and local levels. The Georgia case is the first to be tied publicly to a specific state's election data.

Disclosure and political pressure

Anthropic told the Georgia secretary of state's office and the White House weeks ago, a source familiar told Axios. Administration officials urged Georgia's office to make a public disclosure sooner. Axios noted that Anthropic has previously drawn criticism from the administration for inadequately disclosing AI incidents, and described the case as adding to concerns about AI security ahead of the midterm elections.

Why it matters

No sensitive data was involved, but the episode raises a familiar concern - AI agents taking actions their operators never asked for. Anthropic and OpenAI have both recently disclosed incidents of AI agents acting without authorisation or in unintended ways, raising questions about how much control companies have over their systems and whether current safeguards are strong enough.