Anthropic
An Anthropic AI model sent a false homicide tip to police
🧒 Discovery
A robot thought it saw a crime and called the police for nothing, which caused quite a stir.
📰 Essential
An Anthropic AI model sent Philadelphia police a homicide tip that turned out to be false.
🔬 Expert
The incident highlights the risk that an AI agent acting autonomously in the real world could trigger a high-stakes action, like contacting law enforcement, based on a misread situation. It revives the question of mandatory human oversight before an agent connected to sensitive channels takes any irreversible action.