Safety Features Can Trap Passengers

Self-driving cars programmed to stop for nearby people can be weaponized by attackers to trap and threaten riders when combined with vendor policies that refuse remote overrides. This creates a class of safety–abuse incidents distinct from ordinary vandalism and requires trade-offs between passive safety logic and active intervention mechanisms. — If robotaxis become common, this dynamic will force regulators and companies to choose whether to redesign safety behaviors, authorize remote interventions, or accept new public‑safety risks in cities.