>> Our analysis reveals that risks to people increase with the autonomy of a system: The more control a user cedes to an AI agent, the more risks to people arise. <<

This seems like a "duh".

Risk increases with agency and intelligence.