The legal question surrounding self-driving cars has always been deceptively simple: when the machine is actually driving, who answers when it breaks the rules? China is moving toward a clearer answer. Lawmakers began reviewing a major revision to the country’s Road Traffic Safety Law on August 25, with a dedicated chapter covering autonomous vehicles, road access, traffic violations and insurance.
The most consequential proposal would require vehicle manufacturers or importers to handle road-safety violations committed while an approved autonomous-driving function is activated. That would mark an important shift away from a legal system built around a human sitting behind the wheel. Yet Beijing is drawing a sharp boundary: conventional driving and today’s driver-assistance systems would remain governed by existing rules.
Responsibility Could Move With the Driving Task
The central idea in the draft is that legal responsibility for a traffic violation should follow whoever—or whatever—is performing the driving task. When a vehicle is operating with an autonomous-driving function legally activated, the proposed rules say the manufacturer or importer would accept the handling of road traffic safety violations. That is a significant departure from conventional enforcement, where police normally identify the person driving the vehicle.
China is not treating every technologically advanced car the same way. The draft specifically separates autonomous vehicles from cars equipped with driver-assistance systems. That distinction matters because many vehicles already sold in China can steer, brake, accelerate and navigate under certain conditions without legally becoming autonomous vehicles. The proposal therefore does not give drivers of ordinary assisted-driving cars a broad shield from speeding tickets, signal violations or other offences. The shift applies when the vehicle is operating within the legally recognized autonomous category.
Beijing Is Drawing a Hard Line Between Assistance and Automation
Modern automotive marketing can make the line between assistance and autonomy seem blurry. The proposed law attempts to make that line legally meaningful. If an autonomous-driving function is not activated, the vehicle continues to be managed under rules applying to non-autonomous automobiles. Cars equipped only with assisted-driving technology receive the same treatment, leaving the person behind the wheel responsible for operating them safely.
That distinction is particularly important in China because advanced assistance is already becoming commonplace. Government-linked data released this summer showed that Level 2 combined driver-assistance functions had reached 70.5% penetration among passenger vehicles in 2026, while navigation-on-autopilot functions had reached 34.2%. Those systems can feel highly automated from the driver’s seat, but Level 2 remains fundamentally a driver-assistance category. The proposed legislation reinforces that the driver’s legal role does not disappear merely because software is controlling several functions at once.
The Rules Are Arriving as Intelligent Driving Becomes Mainstream
China is no longer regulating autonomous driving as a distant laboratory experiment. More than 20,000 testing and demonstration permits have been issued around the country, according to government reporting, while more than 57,000 kilometres of roads have been opened for testing and demonstration. Accumulated test mileage has exceeded 220 million kilometres. Twenty cities involved in integrated vehicle-road-cloud pilots have also deployed more than 60,000 pieces of intelligent infrastructure.
Those numbers explain why questions once reserved for engineers are becoming questions for legislators, insurers and traffic police. An experimental prototype operating with technicians aboard can be governed through special permits. Mass-produced vehicles operating alongside commuters require predictable rules about who receives a penalty, what information must be retained and how responsibility changes when control moves between a person and software. The new Road Traffic Safety Law proposal attempts to provide that broader legal foundation while China continues expanding real-world deployments.
China Has Already Put Level 3 Cars on Public Roads
The legislative push follows an important milestone reached in December 2025, when China granted conditional market access to its first two Level 3 passenger vehicles. One came from Changan Automobile and another from BAIC’s Arcfox brand. The approvals moved Level 3 technology beyond conventional testing permits and into regulated road-operation pilots, although only within tightly specified conditions.
The restrictions show how gradual the transition remains. Changan’s approved electric sedan can perform conditional automated driving in a single lane on designated expressways in Chongqing, with a maximum autonomous speed of 50 kilometres per hour in congested conditions. The Arcfox model received authorization for designated expressway sections around Beijing at up to 80 kilometres per hour. Beijing later issued three dedicated Level 3 licence plates. Drivers still have to occupy the driver’s seat as a fallback, illustrating why lawmakers need rules covering both machine control and human takeover obligations.
A Mandatory Safety Standard Is Building the Technical Foundation
China’s legal changes are arriving alongside a new mandatory national safety standard for Level 3 and Level 4 automated-driving systems. GB 44721—2026 was approved on July 30 and is scheduled to take effect on July 1, 2027. It applies to passenger and commercial vehicles equipped with Level 3 or Level 4 systems, while automated parking systems are excluded from its scope.
The standard places extensive responsibilities on manufacturers before a vehicle ever encounters a police camera. Companies are expected to establish safety-management processes covering development, production and post-deployment operation. Simulation, closed-course and public-road validation are required during system development. Regulators also state that an automated-driving system should achieve safety performance at least equivalent to a qualified, attentive human performing the driving task. For Level 3 systems, the vehicle must additionally monitor whether the human fallback driver remains capable of taking control when requested.
A Traffic Ticket Is Not the Same as Accident Liability
One important limitation is easy to lose in a headline. The publicly disclosed proposal addresses the handling of road traffic safety violations when autonomous driving is active. It also establishes that the new autonomous-vehicle chapter will deal with insurance arrangements. The available summaries do not establish a blanket rule saying an automaker automatically pays every civil claim whenever an autonomous vehicle crashes.
That distinction could become critical as the legislation advances. A vehicle might, for example, commit an administrative traffic violation without causing an accident. A collision could also involve several different questions: whether the automated system was active, whether the system malfunctioned, whether a takeover request was issued, whether the human responded appropriately and how insurance contracts allocate compensation. China’s proposed framework starts to relocate responsibility when software is performing the driving task, but more detailed provisions, implementing regulations and eventual court practice will determine how far that principle reaches in individual accident cases.
Vehicle Data Could Become the Equivalent of a Digital Witness
Once responsibility depends on whether automated driving was activated, vehicle records become exceptionally important. A disputed incident may turn on a basic factual question: was the human driving, was a Level 2 assistance feature operating, or had a legally recognized autonomous system taken over the dynamic driving task? That information can determine which regulatory framework applies before investigators even consider what caused the violation.
China’s emerging standards already recognize the importance of data. Safety requirements for combined driver-assistance systems include provisions concerning data recording, while regulators have said they intend to continue developing standards for safety-event data exchange and management. This points toward a future in which logs documenting system status, activation, disengagement and takeover events may become increasingly important evidence. The current public summary of the Road Traffic Safety Law draft does not spell out detailed evidentiary procedures, leaving those mechanics among the major issues to watch.
Automakers Would Gain More Than a Technology Obligation
For manufacturers, autonomous driving is becoming a regulatory responsibility that extends well beyond writing better software. If the proposal takes effect in its current form, an automaker or importer could have to respond when an autonomous vehicle commits a traffic violation. Separately, national standards are already requiring manufacturers to maintain safety processes throughout a vehicle’s life cycle and to demonstrate system performance through multiple forms of testing.
That changes the economics of autonomous driving. A system that performs more of the driving task can create greater convenience for customers, but it can also pull the company deeper into compliance, recordkeeping, insurance and post-sale monitoring. Manufacturers therefore have an incentive to define operating boundaries precisely and make it difficult for autonomous functions to operate where they have not been approved. The same logic explains why current Level 3 pilots restrict activation to designated roads and operating conditions instead of permitting unrestricted use everywhere.
Drivers Should Not Read the Proposal as Permission to Stop Paying Attention
For ordinary motorists, perhaps the most important message is what the draft does not change. Level 2 driving assistance remains assistance. China’s public-security authorities have repeatedly emphasized that drivers using current assistance systems remain responsible for safe vehicle operation and that assisted driving should not be confused with autonomous driving. The proposed law preserves that division instead of treating every hands-off feature as self-driving.
Even Level 3 remains conditional rather than unlimited autonomy. Beijing’s approved Arcfox pilot vehicles still require a human in the driver’s seat who can take over in an emergency. The new mandatory safety standard likewise requires Level 3 systems to monitor takeover readiness. That creates a transitional environment in which responsibility can move between human and machine during the same trip. Clear dashboard warnings, driver education and accurate descriptions of a system’s limitations therefore become legal and safety necessities rather than simply user-experience features.
China Is Helping Write Rules That Could Reach Beyond China
The domestic changes are part of a wider international effort to turn autonomous-driving technology into something governments can regulate consistently. In June 2026, the United Nations World Forum for Harmonization of Vehicle Regulations adopted the first global technical regulation for automated-driving systems. China participated alongside the European Union, Britain, the United States, Canada and Japan and played a major role in developing technical provisions.
China’s new domestic mandatory standard builds on that work while adding requirements tailored to its own roads and regulatory structure. That matters because Chinese automakers are increasingly competing internationally, and autonomous-driving rules can affect which vehicles can be certified and where. The draft Road Traffic Safety Law adds another layer: not simply how an automated system must perform, but who answers when that system violates traffic rules. If enacted, the combination could become an influential model for jurisdictions wrestling with the same transition from human-controlled vehicles to software-driven mobility.
