The rapid evolution of China's intelligent vehicle sector and the widespread adoption of advanced driver-assistance systems are accelerating the real-world deployment of autonomous driving technology. This progress is becoming a cornerstone of the nation's strategy to leapfrog in the global automotive industry. While autonomous driving is poised to become a dominant mode of future transportation, all forms of travel carry inherent risks, raising critical questions: when an accident occurs, who bears the responsibility, and how are disputes resolved? These long-standing industry and public concerns are now on the verge of receiving clear legal clarification.
A draft amendment to the Road Traffic Safety Law, submitted for its first review at the Standing Committee of the 14th National People's Congress on August 25th, specifies that if a road traffic violation occurs while a vehicle's autonomous driving function is activated, the vehicle's manufacturer or importer will be held responsible for handling the matter. This shift in accountability is fundamentally altering the logic of risk distribution. The traditional auto insurance system, long predicated on the principle that "the human is the core of driving," is now confronting a completely new underwriting environment. A pressing challenge for the insurance industry is how to distinguish risk sources under different driving modes and clearly define the risk-sharing mechanism between automakers and vehicle owners.
Automakers to handle violations during autonomous operation.
The draft amendment introduces a dedicated chapter titled "Special Provisions for Autonomous Vehicles," which defines the concepts of autonomous driving and driver-assistance functions. It explicitly states that traffic violations occurring when the autonomous driving function is active will be processed by the vehicle's manufacturer or importer. Vehicles equipped with autonomous driving capabilities that are not activated, as well as vehicles with only driver-assistance features, will continue to be managed under the existing rules for non-autonomous vehicles. This marks a clear legal distinction at the national level between the liabilities associated with autonomous driving and those of driver-assistance systems.
In recent years, China's intelligent vehicle industry has advanced rapidly, with high-level driver assistance becoming mainstream and various autonomous driving scenarios accelerating in deployment. However, the breakneck pace of industrial progress has highlighted the inadequacy of existing legal frameworks. The traditional road traffic management system, built on the principles of manual driving, adheres to a "driver-centric, full responsibility" model. This system functions well for conventional fuel vehicles but is ill-suited for the new paradigm of human-machine co-driving and system-led operation inherent in intelligent vehicles. Should the draft amendment be passed, it will establish a clear entity responsible for handling traffic violations that occur while autonomous driving functions are activated.
Guo Tao, a senior expert in artificial intelligence, noted that this revision addresses a critical gap. Previously, the law treated the driver as the sole responsible party, leaving the question of liability for violations or accidents during autonomous operation without clear legal basis in higher-level legislation. The amendment rectifies this by explicitly distinguishing between autonomous driving and driver assistance, thereby assigning primary responsibility to automakers when their autonomous systems are active.
According to the "Automotive Driving Automation Classification," autonomous driving is divided into levels L0 through L5, representing "emergency assistance," "partial driver assistance," "combined driver assistance," "conditional autonomous driving," "highly autonomous driving," and "fully autonomous driving." Industry insiders point out that if the draft is passed, starting from L3, both driving control and accident liability could, under specific conditions, shift from the human driver to the system itself.
Could spawn two insurance products serving automakers and car owners separately.
With clear liability, supporting services like insurance gain a new reference framework. Cui Dongshu, head of the Passenger Car Market Information Joint Branch of the China Automobile Dealers Association, suggests that the future may see two types of insurance: one purchased by car owners for "non-activated" driving states, and another, a "self-driving product liability insurance," acquired by automakers. Guo Tao echoed this, stating that the draft reserves institutional space for a mandatory insurance scheme specific to autonomous driving, clarifying the boundary of a dual-insurance framework that includes both automaker product liability insurance and owner vehicle insurance.
Regarding product design, Cui Dongshu believes that traditional auto insurance, based on the premise that the "driver is the responsible party," sets premiums using "human factors" such as age, driving experience, and violation history. When a vehicle enters an autonomous driving state, this traditional risk pricing model becomes ineffective. As the draft places responsibility for violations during activated autonomous driving on the automaker, the design logic of insurance products should correspondingly adjust. He draws an analogy to taking a taxi, where both the driver and the platform carry insurance, providing dual protection. The adjustment isn't limited to product design; future auto insurance pricing logic will also evolve. Cui suggests that autonomous vehicles will record extensive driving data, including mileage, scenario complexity, and system takeover frequency, which could lead to fairer pricing where safer driving results in lower premiums and higher risk leads to higher costs.
Information alignment remains a challenge.
Although the draft strictly differentiates between driver assistance and autonomous driving, some vehicles on the market are currently marketed as "L2.99999." During this transitional period before L3 autonomous driving becomes widespread, insurance companies must address the risk exposure created by the gap between automaker marketing claims and consumer understanding. Guo Tao advises that insurance companies should not simply replicate the value-added services offered by automakers. Instead, they should develop standardized intelligent driving add-on insurance products that are filed and regulated. These would supplement traditional auto insurance by covering additional losses stemming from system algorithm defects or sensor anomalies. It is also crucial to clearly define policy exclusions and explicitly inform consumers that even with high-level driver assistance enabled, the driver retains the obligation to take over control of the vehicle.
In terms of risk mitigation, insurance institutions could collaborate with regulators to conduct consumer education, incorporating functional boundaries into prominent policy highlights. They could also implement differentiated premium rates based on the maturity of different vehicles' intelligent driving systems. For models with vague marketing language, risk coefficients could be moderately increased, using price signals to discipline market practices and close the risk gap caused by perception discrepancies.
Beyond assigning responsibility, a more complex challenge in handling autonomous vehicle accidents is accurately determining the vehicle's actual operating state at the time of the incident. For instance, if an autonomous system urgently disengages in the final second before a crash, forcing the driver to take over, how should the associated traffic violation and accident liability be determined? Industry experts note that the underlying operational data, system logs, and operating records of autonomous vehicles are exclusively held by automakers. Car owners, insurance companies, and even third-party agencies often struggle to access this core data, leaving them in a position of information disadvantage during accident evidence collection and liability determination.
Addressing the industry's pain points of difficult evidence collection, challenging liability determination, and information asymmetry, Jiang Han, a senior researcher at Pangoal Institution, proposes several measures. First, regulatory authorities should lead or authorize the establishment of a third-party data platform that uniformly integrates automaker data interfaces, employing an encrypted data-sharing model that is "available but not visible." This would ensure data verifiability while strictly protecting automaker trade secrets and user privacy. Second, for various liability dispute scenarios, rules should be established that prioritize vehicle black-box data as the core basis for evidence and determination. Insurance companies should be granted independent authority to verify original driving data, preventing information asymmetry caused by automakers' unilateral interpretation of data. Furthermore, a professional and independent accident appraisal institution for intelligent connected vehicles should be established. This body would be responsible for analyzing vehicle system logs, using objective technical data as the basis for determining liability, replacing subjective testimony, and effectively safeguarding ordinary consumers' legal rights and equal standing in professional technical disputes.