Waymo robotaxi’s interior AI spots 3‑D‑printed assault rifle, stops vehicle and leads to teen arrests in San Francisco

On 3 September 2024 at around 04:00 h, a Waymo robotaxi operating in San Francisco detected a loaded 3‑D‑printed assault rifle inside the vehicle and brought the car to a stop on the side of the road. The interior camera system, which Waymo uses for compliance monitoring, flagged the firearm as a violation of its terms of use, automatically notified police and the two teenage passengers were arrested shortly thereafter.^heise

Incident timeline and facts

The vehicle’s AI‑driven interior cameras identified the weapon, classified it as an illegal ‘ghost’ assault rifle – a firearm manufactured with a 3‑D printer and lacking a serial number – and triggered the robotaxi’s safety protocol. The car pulled over onto the shoulder, and a police officer conducted a traffic stop described by the San Francisco Police Department as a "high‑risk traffic control".^heise Officers seized the loaded rifle, as well as marijuana and pepper spray found during the search.^heise Both passengers, identified in the report as a male and a female teenager, were taken into custody.

Waymo’s interior monitoring system

Waymo robotaxis are equipped with interior cameras that serve two purposes: passenger safety and compliance with the company’s usage policies. The system runs AI models that can recognise prohibited items, including firearms, in real time. When a violation is detected, the vehicle is programmed to pull over and alert authorities.^heise In this case, the AI correctly identified a 3‑D‑printed assault rifle – a category of weapon the U.S. government has been targeting since 2022 because it cannot be traced through serial numbers.^heise

Waymo confirmed that the stop was caused by a "violation of our terms of use involving a firearm" and that the interior‑camera AI was the trigger for the alert.^heise The company has previously stated that interior monitoring is intended to protect passengers, staff and the public, but it has not disclosed detailed performance metrics for the weapon‑detection model.

Implications for safety, regulation and privacy

The incident demonstrates the practical safety potential of AI‑enabled surveillance in autonomous fleets. By automatically detecting a dangerous object and involving law enforcement, Waymo’s system may prevent incidents that would otherwise rely on human observation. For regulators, the case provides a concrete data point as they consider requirements for onboard monitoring in driverless vehicles.

At the same time, the use of interior cameras raises privacy concerns. The cameras continuously record the passenger compartment, and the AI analyses visual data in real time. While Waymo’s policy states that footage is retained only for safety investigations, the incident shows that the technology can be used to enforce legal compliance beyond traffic safety. The balance between public safety and passenger privacy is likely to become a focal point of future policy discussions.

Waymo’s fleet is expanding rapidly – the company recently launched services in Las Vegas and has announced plans to begin operations in Munich in 2027.^Wikidata The firm employs roughly 2,500 people and is headquartered in Mountain View, California.^Wikidata As the fleet grows, the frequency of interior‑camera alerts may increase, prompting both operational refinements and regulatory scrutiny.

What remains unknown is the false‑positive rate of the weapon‑detection model and how often interior cameras flag non‑threat items. Waymo has not released statistics on the number of alerts generated since the system’s rollout, nor on the proportion that result in police involvement. The company’s public statements confirm the technology works in this instance, but further data would be needed to assess its overall reliability and impact on passenger experience.

For now, the San Francisco incident stands as a rare, verified example of an autonomous vehicle’s AI system intervening in a potential firearms violation. It underscores the dual role of such technology – enhancing safety while introducing new dimensions to the debate over surveillance in public‑facing AI systems.