How is the responsibility for accidents involving intelligent driving of new energy vehicles determined? Beijing pilots the reconstruction of the auto insurance system

Source: Huaxia Times

As intelligent connected technology and new energy vehicles continue to merge, L2 level assisted driving has gradually shifted from a selling point to a standard feature in new cars. High-level autonomous driving at L3 and L4 stages is also moving from testing and verification to practical application.

However, in the process of rapid technological evolution, traditional auto insurance based on human driving logic has fallen behind in the face of “human-machine co-driving” and even system-dominated driving scenarios. Especially with new risks such as hardware/software damage and system failures related to intelligent driving, the existing insurance system not only offers incomplete coverage but also has slow claims processing, making it difficult to meet current practical needs.

Against this backdrop, the Beijing Financial Supervision Bureau took the lead, launching the development and application of commercial insurance products for intelligent connected new energy vehicles. Beijing became the first city in the country to address this industry pain point. From a product design perspective, the intelligent connected new energy vehicle insurance launched in Beijing is not a complete overhaul of the existing system but follows the principle of “overall stability with partial optimization” within the current auto insurance framework, redefining coverage scope and responsibility boundaries. Under this new system, coverage now includes the intelligent driving system itself, including related hardware and software as well as specific functional scenarios. For example, risks during the use of automatic parking, NOA navigation assistance, and other functions, as well as damage to key components like sensors and domain controllers, are all included within the coverage.

Intelligent driving spurs new insurance

Public data shows that by 2025, China’s new energy vehicle penetration rate will surpass 50%. Among them, from January to July, the penetration rate of passenger cars equipped with L2 level assisted functions reached 62.6%, with an expected full-year rate of 66.1%. Regarding high-level intelligent driving, from January to November 2025, cumulative sales of passenger cars equipped with city NOA functions reached 3.13M units, accounting for 15.1% of insured passenger cars, up 5.6 percentage points from 2024. High-level intelligent driving has entered a stage of large-scale popularization.

As the penetration rate of intelligent driving rapidly increases, issues with lagging risk protection systems are also becoming apparent, with numerous real-world examples. For instance, a NIO ES6 owner complained that during the use of automatic parking, the vehicle suddenly lost control and surged forward, crashing into a nearby vehicle. NIO’s after-sales service only provided backend data, claiming the feature was “not activated,” and refused to assume responsibility or compensation; a Xpeng MONA M03 owner reported that while in intelligent assisted driving mode, the vehicle suddenly turned right without slowing down or signaling, then the intelligent driving system shut down on its own, leading to a major traffic accident and vehicle damage. The manufacturer refused to pay for the intelligent driving insurance, citing “human-machine co-driving mode,” and refused to disclose backend data.

These cases highlight the current issue of unclear insurance liability division. In “human-machine co-driving,” traditional auto insurance is built around “human driving,” defining the driver as the natural person with actual control of the vehicle, with the responsibility always on the human. However, once entering the L3 stage and above, the driving process involves both humans and systems, and in some cases, the system even takes the lead. Under these changes, accident responsibility no longer solely corresponds to the driver but may also involve the vehicle manufacturer, software provider, and map service provider, making liability relationships much more complex. Existing auto insurance rules are also difficult to clearly delineate responsibilities among multiple parties.

Automotive industry analyst Pang Wei told Huaxia Times that insurers generally believe that “difficulty in liability determination” is one of the main obstacles in developing insurance for intelligent driving. From actual claims feedback, consumer satisfaction with intelligent driving claims is also significantly lower than that of traditional auto insurance. This unclear liability determination not only affects consumer rights but also increases the uncertainty risk for automakers, to some extent restricting the development and promotion of high-level intelligent driving technology.

Taiping Reinsurance previously released the “White Paper on Innovation in Intelligent Connected Vehicle Insurance,” which pointed out that during the transition to autonomous driving, there is still interaction between the driver and the system, making accident liability determination more complex than in traditional scenarios. This not only raises investigation costs for insurance companies but also prolongs claims processing cycles. In higher-level driverless scenarios, accidents are often caused by network issues, vehicle system failures, or environmental perception problems, shifting liability from the traditional driver to vehicle operators, manufacturers, or technology providers. Insurance products should also adapt to this change, but current market offerings have not fully done so.

Beijing pilot breaks through liability determination difficulties

Regarding current insurance and claims issues for new energy vehicles, Zhang, a manager at Pacific Auto Insurance, told Huaxia Times that overall, new energy vehicles, especially those equipped with intelligent driving features, have premiums about 30% higher than traditional fuel vehicles, with some variation among brands. This mainly relates to the higher hardware costs of the vehicles, such as radars, cameras, and domain controllers, which are more expensive due to their advanced features. In case of an accident, repair and replacement costs also increase, directly affecting premium pricing.

On claims, Manager Wang said, “As long as an accident occurs while the intelligent assisted driving feature is active, and the driver is not involved in illegal activities like drunk driving, unlicensed driving, or drug use, or in cases where the insurance contract explicitly excludes coverage, the claim process can generally proceed. We will not refuse to pay just because the vehicle was in an intelligent driving mode at the time of the accident.”

However, Manager Zhang also admitted that currently, there is no very detailed distinction between “human driving” and “intelligent driving” in insurance products. “Our current approach mainly relies on the accident liability determination issued by traffic police. How they assign responsibility determines our payout proportion. Whether the accident was caused by driver error or a problem with the intelligent driving system, there are no specific standards for liability under the current insurance framework. It’s also difficult for insurance companies to independently determine whether responsibility lies with the human or the system,” Zhang explained.

Regarding the difficulty in liability determination and insurance companies’ challenges in identifying responsible parties, Pang Wei said this is precisely the core difficulty faced by current intelligent driving insurance. The pilot of intelligent driving insurance launched in Beijing aims to solve the disconnect between traditional auto insurance systems and the development of intelligent driving by making adjustments in product design and data to keep insurance services aligned with vehicle intelligence.

To ensure the smooth implementation of dedicated insurance, the Beijing Financial Supervision Bureau stated that the promotion will be carried out in phases and scenarios, following the principle of “maturing one batch, launching one batch.” For the currently most common L2 level assisted driving vehicles, initial products will mainly cover new energy vehicles, allowing owners to choose between dedicated products and existing auto insurance when purchasing. For L3 and L4 level autonomous vehicles, the focus will be on models tested or approved for road use in Beijing, included within the dedicated coverage scope. As data and claims experience accumulate over time, the scope of application will expand gradually, moving from pilot to broader deployment.