Plastic Heads and Printed Photos: How Chinese Drivers Are Bypassing Tesla's Safety Systems

Steering wheel defeat devices are equally low-effort. A water bottle wedged into the wheel, a weighted ring hung from one side, or a dedicated counterweight clip applies continuous torque . The car interprets that pressure as a driver’s hand, preventing the system from issuing repeated “nag” prompts to place hands on the wheel. Some aftermarket sellers market “Nag Elimination Modules”—plug-and-play electronics that partially deactivate the monitoring system outright for around $139 .

Together, these gadgets let a driver take a nap, scroll through a phone, or even leave the driver’s seat while the car operates at Level 2 autonomy—a scenario explicitly prohibited by Tesla’s terms of use .

Why this is exploding now in China

Several forces collided in 2025 and 2026 to create the current surge.

FSD’s cautious rollout in China activated new cabin monitoring. In February 2025, Tesla began pushing an over-the-air update in China that included driver-attentiveness checks via the cabin camera, mirroring features already active in North America . For the first time, a large population of Chinese Tesla owners were subject to constant eye-tracking while using advanced driver-assist features.

A pre-existing gray market for FSD unlocking hardware. Before the camera tricks went mainstream, unauthorized modules costing as little as $140 (a few hundred yuan) were widely sold on Chinese e-commerce sites. These devices plugged into the vehicle’s CAN bus to spoof regional and payment checks, effectively unlocking FSD features without buying Tesla’s official package, which costs roughly $9,000 in China .

Tesla’s 2026 crackdown shut off 100,000 hacked vehicles. In April 2026, reports surfaced that Tesla was remotely detecting and permanently disabling FSD on vehicles fitted with unauthorized CAN bus hardware. Over 100,000 cars were affected, the majority of them in China . Stripped of their hacked FSD capability overnight, drivers turned to a different approach: rather than hacking the software, they would fool the physical sensors and camera.

Cost and availability. Plastic heads cost about $30. Printed photos and cheap LED panels cost even less. All are trivially easy to find on mainstream shopping platforms, placing them well within reach compared to official FSD subscriptions or even the more expensive CAN bus hacks .

Tesla’s monitoring technology has a fundamental blind spot

The gadgets work because Tesla’s driver monitoring relies primarily on a single cabin camera analyzing facial geometry and gaze direction . That camera processes video locally inside the car, and Tesla states no one—including the company—can access the footage remotely . While this protects privacy, it also means the system must make real-time judgments without cloud-based verification.

The problem: a high-quality photo, a plastic head, or a looping LED video presents the same visual pattern as a real, attentive driver. The system checks for “driver’s eyes nominal” and “view of head not truncated”—states easily mimicked by a well-placed replica . There is currently no reliable method built into consumer Teslas that can distinguish a static silicone face from a living human with normal micro-movements, blinking patterns, or skin reflectance changes.

This isn’t a new discovery. As early as 2021, a security researcher demonstrated that a photo taped to the headrest could suppress some driver-monitoring alerts . What’s changed is the scale and commercialization of the trick, fueled by booming FSD demand in China and the sudden removal of software-based workarounds.

Safety consequences are already visible

Circumventing driver monitoring on a Level 2 system has real-world consequences. Tesla’s Autopilot and FSD (Supervised) are not autonomous; they require a human driver able to take over instantly. When a plastic head is the only thing “watching” the road, the actual driver may be fully disengaged.

Concrete examples are emerging:

• In February 2026, a drunk driver with a previous conviction was sentenced to prison after using an aftermarket device to mimic steering wheel contact and bypass attention prompts on a Level 2 system similar to Tesla’s FSD .
• In one widely cited 2023 incident on a Chinese highway, a Tesla driver admitted to sleeping behind the wheel while Autopilot was engaged, later crashing . In online forums, users openly discuss taping over the cabin camera or using sunglasses and visors to blind the eye-tracking system .

These hacks don’t just break terms of service—they fundamentally alter the safety contract between the vehicle and its occupant. The driver-assist system continues to operate under the assumption that a human is supervising, when in fact no one is.

Tesla’s response leaves a persistent gap

Tesla has proven it can fight back—but only against certain types of cheating.

• Steering torque devices: Starting with FSD Beta V10.69.3.1 in late 2022, Tesla introduced code capable of detecting at least some weighted steering wheel defeat devices. If caught, the system can force an Autopilot disengagement and, in the FSD Beta program, issue a strike that could lead to exclusion from the beta .
• CAN bus hacking modules: The April 2026 remote disablement campaign was aggressive and wide-reaching, permanently revoking FSD access for tens of thousands of vehicles .

However, the camera-fooling gadgets present a harder problem. Tesla cannot remotely detect whether a plastic head or printed photo is sitting in the driver’s seat; it can only observe the same video stream the cabin camera sees, which has already been fooled. There is no evidence yet that over-the-air updates have addressed this class of spoof. The cabin camera’s own terms are explicit: using any method to circumvent driver attentiveness monitoring can result in permanent FSD disablement —but enforcement relies on detection capabilities that don’t appear to exist yet for static visual fakes.

Researchers have long warned that camera-only monitoring without robust liveness detection—such as infrared-based eye tracking, depth sensing, or active blink detection—would be vulnerable to simple replay attacks. The current wave of plastic-head tricks in China is a large-scale, real-world validation of that concern .

The wider implication for supervised autonomy

The phenomenon highlights a structural tension in Tesla’s approach. The company is betting that pure vision and AI can handle both external driving and internal monitoring. But cabin-facing cameras equipped only with standard RGB sensors lack the hardware needed to perform the kind of anti-spoofing checks common in modern smartphones or security systems.

As FSD expands into more global markets where regulatory oversight varies, the gap between the system’s capabilities and what users can trick it into doing may grow. The $30 plastic head isn’t a sophisticated exploit—it’s a reminder that supervised autonomy is only as strong as the supervision itself.

Until liveness detection or multi-modal monitoring (such as capacitive steering wheels that sense actual skin contact) becomes standard, driver-monitoring defeats will remain a low-cost, high-stakes loophole—one that turns an attentive-human requirement into an optional setting.