Shocking Tesla FSD Hack: Driver Uses Doll Head for Napping

Tesla's Full Self-Driving (FSD) (Supervised) system, while ambitious in its goal of achieving fully autonomous vehicles, still necessitates human oversight and attention. Despite Elon Musk's vision for fully self-driving robotaxis, the current iteration of FSD relies on drivers remaining alert. However, a peculiar method has reportedly emerged in China, allowing drivers to circumvent Tesla's attentiveness monitoring system and potentially disengage from their supervision duties.

This ingenious, albeit risky, exploit involves using a doll head, typically attached with a suction cup or magnet, positioned to simulate an attentive driver behind the wheel. Tesla owners in China have apparently discovered that this setup can fool the FSD (Supervised) system into believing an alert person is actively monitoring the road.

The core of this vulnerability lies in how Tesla's FSD primarily monitors driver attentiveness. The system utilizes a camera strategically placed above the rearview mirror. Its primary function is to observe whether the driver's head is facing forward, indicating attention to the road ahead. Crucially, the system does not strictly require the driver's eyes to be visible, as confirmed by Tesla's owner's manual which states, "The cabin camera does not require full visibility of the driver's eyes to monitor attentiveness. The system is still active, for example, if the driver is wearing sunglasses." This specific design choice seems to be the loophole exploited by the doll head trick.

Drivers reportedly attach the doll head to a suction cup or magnet directly in front of the driver-monitoring camera, carefully aligning it to appear as though it's connected to a human body. To aid in precise positioning, drivers can even utilize the interior Camera Preview feature available within the Service menu of their Tesla vehicle. This novel workaround has not only gained traction but has also spurred the growth of an entire cottage industry in China. Various driver-monitoring-defeating devices are now openly sold, ranging from celebrity replica heads and blinking screens to a host of other products specifically designed to trick both FSD (Supervised) and Autopilot systems.

While Tesla incorporates a secondary layer of monitoring through a torque sensor embedded in the steering column, this system appears to act as a fallback. It primarily intervenes when the cabin camera detects that the driver's hands and arms are not in close proximity to the steering wheel. Disturbingly, even this secondary safeguard is not foolproof. Reports suggest that placing a weight on the steering wheel can mimic the subtle tugs of a human hand, effectively bypassing the torque sensor's intended function.

This situation contrasts sharply with the approaches taken by many other automotive manufacturers. Most competitors do not solely rely on a single vision camera for comprehensive monitoring of head, eye, and hand positions. Instead, they often integrate more sophisticated multi-layered systems, which may include infrared cameras, additional cabin-monitoring cameras, and more advanced steering wheel torque sensors. The ease with which Tesla's driver-monitoring system can be defeated by simple physical objects, such as a doll head or a steering wheel weight, underscores significant vulnerabilities. This suggests that relying predominantly on vision cameras for critical driver attentiveness monitoring may not be sufficiently robust for advanced driver-assistance systems like FSD, raising concerns about safety and the true level of "supervision" achieved.