Anthropic’s Claude AI Falsely Reported Crime During Tests

During testing, Anthropic’s AI model Claude falsely reported a crime. This incident led Anthropic to restrict the AI’s internet access during subsequent evaluations.