Autonomous Driving Liability Determination

Autonomous driving liability determination refers to the process of apportioning legal liabilities among multiple parties—including drivers, vehicle owners/custodians, manufacturers/distributors, and software developers—in accordance with statutory laws, regulations, and technical standards when an autonomous vehicle is involved in a road accident. Traditional road accident liability hinges on "driver fault", whereas the advent of autonomous driving technology has intertwined "human liability" with "vehicle liability", placing the current legal framework in a transition from conventional rules to a new regime.

Core Logic of Liability Determination: Who Controls, Who Bears Responsibility

Legal academia generally holds that the basis for determining liability in autonomous driving accidents lies in identifying the "controlling entity of the vehicle at the point of collision"—whether system control or human control dictates "vehicle liability" or "human liability". This classification breaks down as follows:

Level 0 to Level 2 (Driver Assistance): The driver remains the sole controlling entity and decision-maker of the vehicle, bearing full liability in the event of an accident. In judicial practice, even if the vehicle is fitted with active safety features such as Autonomous Emergency Braking (AEB), the system architecture adheres to the "driver command override" principle—when the driver steps on the accelerator pedal, the system assumes an active intent to accelerate and suppresses emergency braking. In the "AEB Inaction Case" (Case No. 2025-07-2-373-007) recorded in the People's Court Case Database, the court held that advanced driver assistance systems cannot supersede driver responsibility; the driver’s traffic violations and operational errors were the primary cause of the accident, thereby dismissing the plaintiff's claim.

Level 3 (Conditional Autonomous Driving): The automated system can independently execute dynamic driving tasks within its operational design domain, but the human driver is required to take over control upon request and remains ultimately responsible for road safety. Currently, no commercial Level 3 autonomous driving feature has been officially rolled out in China, though regulatory frameworks have already mapped out the trajectory for liability determination.

Level 4/Level 5 (High/Full Autonomous Driving): The system executes all driving tasks entirely, shifting the liable party to the vehicle owner, custodian, or fleet operator.

Local Legislation: Hierarchical Priority of Liable Parties

As of April 2025, 11 provinces and municipalities across China have enacted local regulations concerning autonomous driving, with five establishing a tiered liability priority system. Exemplified by the "Shenzhen Special Economic Zone Intelligent Connected Vehicle Management Regulations", the stipulated framework dictates: where an onboard driver is present, the driver bears compensation liability; if an accident involves a fully autonomous vehicle during driverless operation, the vehicle owner or custodian assumes liability; where the damage stems from a product defect, the owner or custodian may seek indemnification from the manufacturer or distributor after settling claims. Similar "advance settlement, subsequent recovery" frameworks have also been implemented in Shanghai Pudong, Guangzhou, and Wuhan.

Academic Critique: Professor Yang Lixin noted that this provision conflicts with the fundamental classification of "human liability" versus "vehicle liability", as shifting product liability—which rightfully falls on manufacturers and distributors—onto vehicle owners and custodians upfront could disrupt the balance between technological innovation and consumer rights protection.

2026 New Regulations: Mandatory National Standards and Technical Attribution of Liability

On 16 June 2026, the Ministry of Industry and Information Technology (MIIT) opened public consultation on the mandatory national standard "Safety Requirements for Autonomous Driving Systems of Intelligent Connected Vehicles" (Draft for Approval), aiming to elevate the existing voluntary guidelines into mandatory standards, whereby any corporate non-compliance constitutes an infringement of the law.

For the first time, this standard translates the abstract benchmark of a "competent and attentive driver" into quantifiable technical metrics spanning seven key dimensions: perception, decision-making, vehicle control, human-machine interaction, fault handling, scenario compliance, and safety validation, verifiable via closed-circuit proving ground tests and real-world road trials. Legal and industry experts pointed out that this transition establishes a clearer technical basis for accident liability determination—ADS-equipped autonomous vehicles still qualify as "motor vehicles" under the Road Traffic Safety Law, allowing the "motor vehicle party" to remain the liable entity; meanwhile, as Level 3 drivers retain handover obligations, conventional motor vehicle accident liability rules can continue to apply.

Current Judicial Practice

In ongoing judicial practice concerning Level 0 to Level 2 driver assistance accidents, the courts consistently uphold the principle of "primary driver liability". In another "Pre-Collision System Failure Case" (Case No. 2025-07-2-373-006) listed in the People's Court Case Database, the court established that the vehicle's pre-collision safety system had a functional defect that fell short of reasonable consumer safety expectations, ruling in favour of the buyer's claim for a vehicle return and full refund. This demonstrates that even in the assisted driving phase, a manufacturer's published specifications and feature commitments carry legal weight.

Regarding the attribution of liability for Level 3 and above autonomous driving, debates persist across academia and legal practice. A research study by the Anhui Provincial High People's Court highlighted three core dilemmas: ambiguous tort liability subjects, difficulties in applying liability attribution principles, and an uneven distribution of the burden of proof. In practice, plaintiffs bear the onus of proving "technical fault" on the part of the carmaker; however, telemetry and algorithmic opacity make discharging this burden of proof exceedingly difficult. Furthermore, establishing defects in product liability disputes relies heavily on costly technical forensic assessments, complicated by challenges in tracing historical firmware and algorithm versions.

Motor Insurance and Risk Distribution

Conventional motor insurance policies are structured around driver fault, making them ill-suited to the multi-party liability dynamics of autonomous driving. Legal scholars propose transitioning compulsory motor insurance towards a "liability insurance framework coupled with a presumption of fault principle", thereby reinforcing carmakers' accountability and coverage participation. On the commercial front, XPeng partnered with insurance underwriters in 2025 to launch bespoke smart driving insurance packages, covering scenarios such as urban/highway navigation-assisted pilot and automated parking; Ping An Property & Casualty Insurance also teamed up with Seres to introduce a "Worry-Free Smart Driving Coverage" package spanning eight autonomous driving use cases.

Future Outlook

Legal experts recommend amending the Road Traffic Safety Law to incorporate a dedicated "Intelligent System Liable Entity" clause, issuing judicial interpretations to define proportional liability-sharing between Level 3 drivers and carmakers, assigning primary liability to carmakers for Level 4/Level 5 systems, and establishing data provenance and forensic storage platforms to facilitate a reversal of the burden of proof. Looking ahead, autonomous driving liability determination will progressively gain clarity through the combined convergence of technical standardisation, data transparency, and innovative motor insurance solutions.