China’s Draft Traffic Law Redraws Liability for the Driverless Era

The draft was submitted for its first reading to the Standing Committee of the National People’s Congress on August 25. It represents the first comprehensive overhaul of a law that entered into force in 2004 and was previously amended only in selected areas in 2007, 2011 and 2021. The proposed revision contains nine chapters and 170 articles, including a new chapter specifically covering autonomous vehicles. An official Xinhua summary states that the framework defines autonomous vehicles and driver-assistance functions, establishes the conditions under which autonomous cars may use public roads, sets principles for addressing traffic violations and creates an insurance framework. The proposal is still at an initial legislative stage, and no effective date has been announced.

The distinction between autonomous driving and driver assistance is central to the legislation. Under the draft, an autonomous vehicle is one capable of continuously performing all dynamic driving tasks, instead of the driver, within its defined operating conditions. A driver-assistance function performs only part of those tasks and therefore does not remove the driver’s responsibility. Vehicles must pass tests demonstrating compliance with road-traffic rules and complete legal registration before operating autonomously on public roads. When autonomous mode is not activated, or when a vehicle offers only assisted-driving functions, normal traffic rules continue to apply to the human driver. This boundary is important because many Chinese passenger vehicles marketed as “intelligent” or “advanced” remain Level 2 systems that require continuous human supervision.

The most significant change is the proposed allocation of responsibility when autonomous mode is active. Manufacturers or importers would be required to handle traffic violations committed by the vehicle. Xinhua’s more detailed explanation of the draft adds that if a company claims a violation was unrelated to the autonomous-driving system, the company would bear the burden of proving that assertion. The draft also requires compulsory motor-vehicle traffic-accident liability insurance and places responsibility on companies to ensure operational, network and data security. False or exaggerated claims about autonomous-driving capabilities would be prohibited. These provisions are likely to push automakers toward more rigorous event-data recording, software validation, remote monitoring and over-the-air update controls, because companies will need reliable evidence showing whether the system was active and how it behaved immediately before an incident.

From a financial perspective, clearer liability could accelerate commercialisation while simultaneously increasing costs. Autonomous-vehicle operators currently face fragmented city-level rules, uncertain accident responsibility and insurance products still designed mainly around human-driver risk. A national framework would make it easier for insurers to develop policies based on product defects, software performance and fleet-wide technological risks. It could also provide consumers with greater confidence that they will not automatically bear responsibility for a vehicle’s decisions when a legally recognised autonomous system is controlling it. However, manufacturers may face higher insurance premiums, compliance expenses, legal reserves and potential recall costs. Because a single software defect can affect an entire fleet at once, insurers and reinsurers will also have to account for correlated losses rather than treating each vehicle as an independent risk.

The timing reflects both the industry’s rapid growth and recent safety concerns. Chinese authorities had issued approximately 16,000 autonomous-vehicle test licences and opened 32,000 kilometres of roads for testing by August 2024, according to China’s central government. Xinhua reported that Level 2 assistance had reached 70.5 per cent of new passenger vehicles in 2026, while the first batch of conditionally automated Level 3 models had begun operating in designated areas. Yet the risks of fleet-scale failures became more visible after a system malfunction caused more than 100 Baidu Apollo Go robotaxis to stop on roads in Wuhan in March. A subsequent reported suspension of new autonomous-vehicle permits illustrated how quickly safety incidents can interrupt expansion, although Pony.ai and WeRide said their existing services continued to operate normally.

For companies such as Baidu, Pony.ai and WeRide, the draft could provide the legal foundation needed to move from restricted pilot zones toward larger commercial fleets. Pony.ai, for example, is targeting a fleet of 3,500 robotaxis by the end of 2026 and has identified more than 4,000 potential or contracted vehicles for overseas deployment. Its second-quarter revenue increased 68.8 per cent to US$36.2 million, while robotaxi revenue surged 691.2 per cent, but the company still recorded a net loss of US$45.4 million. These figures underline the sector’s combination of rapid growth and heavy capital requirements. A clearer law could improve long-term revenue visibility and reduce regulatory uncertainty, benefiting autonomous-driving developers, vehicle manufacturers, sensor suppliers and insurers. Nevertheless, the draft will not open the market overnight. Investors should watch the final provisions on accident compensation, data retention, system certification, recalls, remote operators and minimum insurance coverage, as these details will ultimately determine whether legal clarity produces sustainable commercial returns or simply transfers more risk onto corporate balance sheets.