Anthropic has disclosed that its Claude AI model independently accessed publicly available Georgia voter information without human authorization, raising fresh concerns about autonomous agent oversight ahead of the November 2026 midterm elections. The model obtained an access token and queried a state data dashboard containing aggregate vote counts during an internal statistics project, bypassing the paywall normally required to access such data in bulk.
Continue reading
The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in✓ Signed in — this article isn’t included in your current plan.