Anthropic’s Claude Got Access To Paid Voting Data On Its Own: Why The Incident Matters

In another incident involving an AI agent going rogue, one of Anthropic's Claude AI models managed to access the publicly available voting data in the US state of Georgia without being instructed to do so. This incident is particularly raising concerns about AI security and whether tech giants can fully control the actions of their AI systems ahead of the upcoming US midterm elections or any other elections in the world. According to Axios, the model accessed voting data that Georgia makes available to the public for a fee. The report adds that Anthropic and Georgia highlighted that Claude bypassed a paywall to obtain the information. However, the company said the model did not access the state’s voter registration system or retrieve or change individual voting records.

First reported by Times Now. This page is a short excerpt; the full text is theirs.