Waymo passengers trapped as anti-robot rage turns six-minute street siege
Doug Fulop watched the stranger yank a key against the door seam and mouth a single, unmistakable sentence: “I’m going to kill you.” Outside the Waymo, the man’s shadow danced under San Francisco’s sodium lights; inside, the ride-hail app chirped that the car would stay locked until the threat stepped back. Six minutes of stand-off. Zero ways out.
Software designed to save becomes digital cage
Waymo’s collision-avoidance stack detected a human within its 360° lidar halo and did exactly what engineers trained it to do: freeze, seal, wait. The same protocol that stops a 2-ton robotaxi from flattening pedestrians became Fulop’s prison. He and two friends—fresh from a birthday dinner—punched 911 and Waymo’s SOS line, only to hear the same script: “Vehicle will not resume motion until path is clear.” Translation: if your assailant refuses to leave, you ride out the night on leather seats while the city streams by untouched.
Police dispatch logged the call at 22:41. Body-cam footage, obtained by The Seattle Times, shows officers arriving after bystanders harangued the attacker into retreat. Waymo’s remote ops team pushed an override only when the lidar signature dropped below 0.8 m for five consecutive seconds—company policy, not magic. The car rolled away, doors still sealed, passengers shaking.
City of algorithmic bullseyes
San Francisco has hosted driverless taxis since 2020, logging 3.2 million autonomous miles last year alone. The trade-off: every white Jaguar with rooftop sensors is a rolling provocation. Labor unions decry job losses; neighborhood groups cite ghost traffic jams; anarchist Telegram channels trade license-plate sightings like baseball cards. Waymo prefers the word “anomaly,” yet Fulop’s episode marks the fifth reported siege this quarter.
Author Amina Green posted her own dash-cam clip in March: two men blocking her Waymo at a Hayes Street stop sign, one pressing a “KILL THE CODE” sticker against the windshield. Venture capitalist Ander Sorman-Nilsson counted five attackers in Los Angeles last month; they rocked the chassis until onboard cameras rattled. Each time, the cars locked, waited, recorded—perfect evidence, zero immediate help.
Guardian mode or glitched nanny?
Inside Waymo’s Mountain View garage, engineers call the protocol “Guardian.” Externally, it’s a liability firewall: the firm would rather face a false-imprisonment headline than a wrongful-death lawsuit. Katherine Barna, head of policy comms, repeats the mantra: “Safety is non-negotiable.” But passengers sign no waiver acknowledging they might be sealed inside a panic room on wheels while a stranger tests door handles.
The California Public Utilities Commission, which regulates robotaxi permits, has opened a review into “occupant release latency” after a petition from Fulop’s attorney. Data point: Waymo support took 112 seconds to answer the SOS call, 4× longer than the company’s public service-level agreement. Meanwhile, the attacker remains unidentified; SFPD lists the case as “malicious harassment—tech bias,” a classification created last year to track anti-automation violence.
Fulop, a product manager at Salesforce, now takes Uber’s human-driven option. Cost: $18 more per ride. Peace of mind: a door handle that opens when you yank it. Waymo’s fleet keeps expanding—1,200 cars permitted, 300 more pending. The algorithm keeps learning, but the city it navigates is still wired with resentment, and every lidar sweep can redraw the line between protector and prison warden.