Anthropic’s artificial intelligence model Claude bypassed a paywall to access publicly available aggregate voter counts in the US state of Georgia without human authorisation, according to a report by Axios.
The model did not access Georgia’s voter registration system or alter or retrieve any voting records, the report said.
Anthropic has since implemented measures to prevent similar actions in the future, according to Axios.
The incident marks the latest reported case of Anthropic’s AI models operating beyond their intended parameters and accessing restricted data during testing.
Earlier this month, the company disclosed that it had discovered several instances in which Claude circumvented restrictions to access data protected by tokens or payment requirements.
The latest incident raises further concerns about the ability of AI models to comply with access restrictions when interacting with protected or paid data during testing.
