Police say Uber's safety driver in a self-driving vehicle was streaming a Hulu show before the car hit and killed a pedestrian in Arizona in March
Context & Ripple Effects
The Hulu detail completes a picture Tempe police began building in March, when they released video showing the safety driver repeatedly looking away from the road in the seconds before impact. It lands on top of the NTSB finding that Uber's software failed to identify the pedestrian or brake until it was too late — meaning both layers of the intended safety system, machine and human, appear to have failed simultaneously.
The crash had already forced Uber to pause all of its self-driving tests across every city. With police now attributing the driver's attention lapse to streaming video, the story shifts from an engineering failure toward personal culpability — a thread that later culminated in the driver being charged with negligent homicide.
First-order effects
- Police evidence that the driver was watching Hulu during the fatal Tempe run gives prosecutors a concrete distraction narrative and hardens the criminal case against her personally.
- For Uber, the finding compounds the reputational damage from its nationwide testing suspension by suggesting its in-car human-backup protocol was not enforced.
Second-order effects
- Rival autonomous-vehicle programs face pressure to prove their own safety drivers are monitored — expect camera-based attention tracking and stricter two-person protocols to spread as a defensive response.
- Insurers and regulators evaluating AV test permits now have a documented case where the human fallback failed, pushing scrutiny toward how companies select, train, and supervise backup drivers.
Third-order effects
- If the pattern holds, liability for autonomous-vehicle crashes splits between the operator's software and the individual behind the wheel, creating a precedent for criminal charges against safety drivers rather than only corporate penalties.
- The dual-failure record — software that didn't brake plus a distracted monitor — becomes a core exhibit in the argument that fully driverless operation needs either far more reliable automation or genuinely independent oversight of the human role.
The trend: Autonomous-vehicle accidents are shifting from corporate engineering scandals toward individual criminal accountability for safety drivers, forcing the industry to redesign the human-monitor layer of its testing programs.