Your Robotaxi Could Be Spying on You

In early September, two teens hopped into a Waymo robotaxi, only to find themselves in the back of a police car shortly after. The autonomous vehicle company reported that it detected a “violation of our terms of service involving a firearm,” prompting it to pull over and notify emergency services. Police later arrested the passengers when they allegedly discovered a loaded AR-style “ghost gun” inside the vehicle.

This incident has sparked a flurry of questions regarding privacy and the use of autonomous vehicles, particularly concerning what these vehicles record, how that footage is handled, and the level of privacy riders can expect during their journeys. Waymo’s fleet, equipped with numerous cameras, has already been at the center of privacy debates. Law enforcement has previously obtained search warrants for footage captured by Waymo vehicles, including a case last year where the LAPD used footage from a Waymo to investigate a hit-and-run. During protests in Los Angeles, several Waymo vehicles were vandalized amid concerns that they might be recording protesters for potential identification or arrest. Waymo has not disclosed how many requests it receives from law enforcement or how often it complies.

These worries echo the backlash against Flock’s network of automated license-plate readers, which have raised alarms about government surveillance and misuse. Unlike stationary Flock cameras, Waymo vehicles navigate through cities, equipped with cameras both inside and outside.

The recent case involving the arrested teens escalates these concerns. In this instance, police were not seeking footage after a crime; instead, Waymo itself detected something suspicious and initiated the police response.

A Waymo spokesperson explained that the company utilizes its internal cameras to enforce its terms of service, including identifying prohibited items and behaviors. The spokesperson noted that automated technology helps identify potential violations, such as firearms. “If Waymo detects a firearm in the vehicle, we may reach out to local law enforcement,” they added.

However, Waymo did not clarify how its automated system identifies firearms, the confidence level required to flag an item, the role of human operators in the process, or what other violations might trigger a police response. These omissions are significant. This situation was not a human driver making a judgment call; it was a driverless car, equipped with cameras and AI, escalating a situation involving its passengers to law enforcement.

Bob Somers, a former engineer at an autonomous vehicle company, suggested that Waymo’s system likely scans for various dangerous items and behaviors. The software might not need to confirm the presence of a firearm definitively; it could flag something that resembles one for further investigation.

“If there is a certain set of objects detected, then that would probably loop in a remote operator to take a look at what’s going on,” Somers explained.

Phil Koopman, a professor emeritus at Carnegie Mellon University who studies autonomous vehicle safety, noted that computer vision systems can be trained to recognize objects and potentially hazardous situations. “If you can train a computer to look at the road and decide how to drive, you can train a computer to decide if somebody’s holding a gun,” he said, though he cautioned that these systems are “never 100 percent accurate.”

This raises the critical question of human review. A human taxi driver has context; they’ve been present for the entire ride, can hear conversations, and can assess whether an object is a real weapon, a toy, or something else. Earlier this year, Waymo reported two teens to the police for underage drinking and shooting a toy gun filled with water beads from the vehicle, though it remains unclear if any charges were filed.

A remote reviewer may lack this context, as they haven’t experienced the ride firsthand. Waymo states that its interior microphones are typically active only during calls with Rider Support or when riders enable them, meaning a reviewer might not hear what led up to a flagged moment. Currently, Waymo employs around 70 “remote assistance agents” in the U.S. and the Philippines, who provide support to the Waymo Driver but do not control the vehicle. It’s unclear if those reviewing potential violations are among these agents.

“It’s really important that Waymo gets it right when they alert the police to something dangerous,” Koopman emphasized. A false alarm could not only undermine riders’ trust in robotaxis but also put passengers at risk during police encounters.

Additionally, there’s the issue of what Waymo communicates to riders about their expectations. Without a human driver, robotaxis might feel more private than they actually are. In 2023, some individuals reported engaging in intimate activities in Cruise robotaxis, despite the presence of cameras. While no similar reports have surfaced regarding Waymo, these instances highlight how easily passengers might confuse the absence of a human driver with a lack of oversight.

Waymo’s terms of service require riders to comply with laws and prohibit dangerous conditions, weapons, or drug use. Riders must be at least 18 to ride alone in San Francisco, and the two arrested teens allegedly violated both age and weapons restrictions. The terms state that Waymo can suspend or terminate a rider’s access if it believes they have violated the rules. Earlier this year, Waymo suspended an account after three boys were photographed hanging out of the vehicle’s windows. However, the terms do not explicitly inform passengers that the company may contact law enforcement based on what it observes inside the car.

Waymo’s public camera policy indicates that its interior cameras can be used to ensure compliance with in-car rules and promote safety. It states that Waymo Support may review recorded video after an issue arises and, in urgent situations, access live video during a trip. The policy also mentions that Waymo may share certain data with law enforcement as needed, but it does not specify that automated monitoring could lead to Waymo contacting police.

Hayley Tsukayama, director of state affairs at the Electronic Frontier Foundation, criticized Waymo’s policies as “very broad.” She argued that riders shouldn’t have to infer from vague language about “safety and security” that their actions could result in police involvement, insisting that Waymo should clearly disclose this possibility before passengers enter the vehicle. “I’m not condoning illegal behavior in rideshares,” she stated, but “if they’re going to share it with authorities, then that should be clearly stated in the policy.” As it stands, Tsukayama believes this is “outside of the expectations of consumers.”

As robotaxis become more prevalent, riders may need to adjust their expectations regarding what occurs inside these vehicles. A human driver can make decisions based on their experiences during a ride, while in a robotaxi, cameras and software can flag incidents, remote workers may intervene, and the vehicle can autonomously summon police.

While Waymo’s system may have assisted in recovering an illegal firearm in this case, it also raises fundamental questions about the nature of a robotaxi: is it merely a ride with safety cameras, a continuously monitored environment, or a system capable of reporting its passengers to law enforcement?

This theme is already emerging in fiction, with robotaxis appearing as antagonists in upcoming narratives, such as a short film about a driverless car that turns against humans. One unsettling concept is that a robotaxi could evolve from a mode of transportation into a form of law enforcement. This month provided a glimpse of what that future might entail.