A Waymo passenger bolted from a self-driving car after the vehicle froze on Phoenix light-rail tracks with a train bearing down, newly released video shows.
At a Glance
- A Waymo stopped on Phoenix light-rail tracks Jan. 7 as a train approached
- Passenger fled the vehicle; trains reversed to avoid collision
- Scene cleared in about 15 minutes with no injuries or service disruption
- Why it matters: The incident highlights rare but serious edge-case failures as autonomous cars expand across cities
The footage, captured by a bystander near Central and Southern avenues in south Phoenix, shows the rider sprinting away while the empty car inches forward along the tracks. Valley Metro staff spotted the vehicle, halted approaching trains and reversed them to create distance. The extension where the event occurred opened only months ago and remains unfamiliar to many GPS maps.
Construction Confused the Car
Arizona State University professor Andrew Maynard, who reviewed the video for KPNX, called the situation a textbook “edge case.” He said construction barricades and the recently laid track likely scrambled the car’s sensors.
- New rail lines often lack detailed LiDAR mapping
- Temporary cones and barriers can mask track edges
- Autonomous fleets rely on pre-loaded 3-D maps updated weekly
“Humans are really good at seeing a brand-new situation and trying to work out how to get around it,” Maynard told the outlet. “These vehicles are smarter than humans when they know the road, but they can falter when something completely unexpected appears.”
Phoenix police confirmed they received a 911 call around 9:15 a.m. By the time officers arrived, the Waymo had already driven itself off the right-of-way. Rail service resumed within 15 minutes; no injuries or property damage were reported.
Pattern of Close Calls
The Jan. 7 incident follows another narrowly avoided crash captured on dashcam last fall. In that footage, a Waymo suddenly veered into oncoming traffic, forcing an approaching driver to swerve and honk. The passenger, a first-time rider, can be heard screaming as the car continues for several seconds before correcting itself.
Sophia A. Reynolds reported the rider felt “helpless” and unsure whether to grab the wheel or wait for the system to respond. The company later blamed a misclassified lane marking.
Safety Record Under Scrutiny
Waymo advertises its vehicles as having fewer distractions and faster reaction times than human drivers. Federal data show autonomous test fleets in Arizona logged more than 3.8 million driverless miles in 2025 with only a handful of injury crashes. Still, edge cases-rare, high-risk scenarios-remain the technology’s Achilles’ heel.
| Metric | Waymo 2025 | Human-Driven Cars |
|---|---|---|
| Crashes per million miles | 0.4 | 2.0 |
| Injury crashes per million miles | 0.08 | 0.78 |
| Edge-case disengagements | 1 per 11,000 mi | N/A |
Maynard stressed that while the Phoenix video looks dramatic, autonomous systems usually err on the side of stopping completely rather than risking the wrong maneuver. “The car effectively chose paralysis over the chance of hitting something,” he said.
What Happens Next
Valley Metro has asked the city to add temporary curb extensions and clearer track markings while construction wraps up. Waymo, which has not yet responded to News Of Losangeles‘s request for comment, typically updates fleet software within days when mapping gaps surface.
Passengers who encounter similar situations can:
- Press the Help button inside the vehicle
- Use the mobile app to summon a remote operator
- Exit the car if safe and call 911
Phoenix police encourage riders to remain inside with seat belts fastened unless there is an immediate threat, noting that stepping onto live tracks carries greater danger than staying put.
Key Takeaways
- Autonomous cars rely on highly detailed maps; new rail lines can pose challenges until updated
- Edge cases-rare, unexpected scenarios-still trip up even the most advanced systems
- Quick action by Valley Metro staff prevented what could have been a catastrophic collision
- No federal rules require public reporting of near-miss events, making real-world footage critical for transparency
