Drivers in China using plastic head figurines to bypass Tesla autopilot safety monitoring

**The Cheap Workaround Defeating Tesla's Driver Monitoring**

A thriving cottage industry has emerged in China selling small plastic figurines designed to fool Tesla's cabin-facing cameras into registering an attentive human driver. The products sell for as little as $30 and come in novelty forms — including replicas of celebrity faces, among them a miniature head styled after Dwayne Johnson — meant to be mounted above the rearview mirror in the sightline of Tesla's interior monitoring camera. Because Tesla's Autopilot and Full Self-Driving systems require the driver to demonstrate attention by tracking eye and head position, the figurines effectively satisfy that check without a human doing anything at all. The actual occupant is then free to use their phone, nap, or disengage from the driving task entirely, all while the vehicle continues to operate in assisted-driving mode.

The products have spread rapidly across Chinese e-commerce platforms, where sellers market them openly alongside other accessories. Beyond static plastic heads, the workaround ecosystem includes small screens that display looping video of blinking eyes and moving heads — designed to pass as a live person to the camera's motion-detection logic — as well as other DIY configurations shared in online communities. The fact that multiple competing product types exist at low price points suggests the market has been active long enough for iteration and competition to take hold. It is not a single obscure hack but a recognized consumer category with multiple vendors.

The core vulnerability being exploited is Tesla's reliance on computer vision to infer driver attentiveness. The system is apparently unable to distinguish between a three-dimensional human face and a sufficiently realistic proxy object placed at the correct distance and angle. This is a meaningful gap: the entire safety premise of semi-autonomous driving features is that a human remains ready to intervene if the automation fails. When that monitoring layer can be defeated with a $30 novelty item, the redundancy disappears entirely.

**Why China Is the Focal Point for This Behavior**

China is now one of Tesla's largest and most strategically critical markets, and it is also where the company has been aggressively rolling out more advanced assisted-driving features to compete with domestic automakers like BYD, Huawei-backed Aito, and Xpeng, all of which have launched their own driver-assistance systems. The competitive pressure has pushed Tesla to expand the availability of Full Self-Driving capabilities in China, meaning more vehicles on Chinese roads are operating with the camera-based driver monitoring enabled.

Chinese drivers have also demonstrated particular enthusiasm — and a particular pragmatism — in pushing assisted-driving systems to their operational limits. Social media platforms like Douyin and Bilibili are filled with videos of drivers testing or showing off semi-autonomous features in ways that would alarm regulators in the United States or Europe. The cultural framing often treats these experiments as clever or impressive rather than dangerous, and the online sharing of workarounds accelerates adoption across a wide user base. The figurine market feeds directly into that ecosystem: a technique discovered by one driver becomes a product sold to thousands within weeks.

There is also a regulatory dimension. China's rules around distracted driving and autonomous-vehicle oversight are still evolving, and enforcement of in-cabin behavior is difficult. Unlike some jurisdictions that have installed roadside cameras capable of detecting phone use, monitoring what a driver is doing based on whether a plastic head is in frame is not a challenge Chinese traffic authorities are currently equipped to address. That regulatory gap creates space for the workaround market to operate without immediate legal consequence.

**What This Reveals About Semi-Autonomous Safety Architecture**

The figurine phenomenon exposes a structural weakness in how current Level 2 driver-assistance systems — the category that includes Tesla's Autopilot and FSD — are designed and validated. These systems are legally and technically defined as requiring continuous human supervision. The automaker satisfies its obligation to enforce that requirement through the driver-monitoring camera. But the camera's effectiveness depends entirely on its ability to accurately classify what it sees, and that classification can be spoofed.

Tesla is not alone in this vulnerability. Nearly every automaker deploying camera-based driver monitoring faces the same fundamental challenge: the system has to distinguish genuine human attention from simulation, and as these products demonstrate, simulation can be surprisingly low-tech. More sophisticated monitoring approaches — including infrared eye-tracking, weight sensors in the seat, grip-detection on the steering wheel, or biometric inputs — exist and are used in varying combinations by different manufacturers, but no system has proven fully tamper-proof.

The deeper implication is about the gap between the legal framework governing semi-autonomous driving and the behavioral reality on the road. Regulators in the United States, Europe, and China have generally accepted that Level 2 systems can be deployed safely because human oversight is nominally maintained. The figurine workaround is a vivid, concrete demonstration that nominal oversight and actual oversight are not the same thing. When the monitoring mechanism is a camera that cannot tell Dwayne Johnson's plastic likeness from a living person, the safety case rests on an assumption that determined or careless drivers will choose not to undermine it — which is not an assumption that holds under real-world conditions.

For Tesla specifically, the China figurine market creates a reputational and legal exposure problem. Any serious accident involving a driver who had bypassed monitoring using one of these devices would raise immediate questions about whether the system's safeguards were adequate and whether the company had taken sufficient steps to detect and block known circumvention methods. Whether Tesla has the technical capability to update its detection algorithms to recognize static or looping-video proxies — and how quickly it can do so given the pace of workaround product iteration — will likely determine how long this particular vulnerability remains a practical risk.