China Issues First Mandatory National Standard for L3/L4 Autonomous Driving, Effective July 2027

The national standard "Intelligent Connected Vehicles—Safety Requirements for Autonomous Driving Systems" (GB 44721-2026) has been officially released for public review. This is China's first mandatory national standard addressing the safety requirements of L3/L4 intelligent connected vehicle autonomous driving systems, and it will take effect on July 1, 2027. All models equipped with high-level autonomous driving features currently on sale must subsequently comply with the standard. It defines general technical requirements for autonomous driving systems in dimensions such as dynamic driving task execution, human-machine interaction, and user notification, and clarifies vehicle manufacturers' safety assurance obligations and the corresponding test admission conditions for functions.

For L3 autonomous driving functions, the standard specifies technical red lines for takeover capability monitoring, detailing three mandatory requirements: seat belt status monitoring, driver presence monitoring, and dynamic driving task execution capability monitoring before takeover. For L4 autonomous driving functions, the standard establishes dedicated technical provisions, offering implementation guidance on dynamic driving task execution logic, background remote assistance rules, and human-machine interaction logic in special scenarios. The standard also requires that the entire process of activation and deactivation of autonomous driving systems be free of security vulnerabilities, and that prompt signals for different states—such as system ready, normal activation, and imminent exit—be clear and easy for ordinary drivers to understand.

The standard requires vehicle manufacturers to build safety assurance capabilities across the full lifecycle, from product design and development, through production and manufacturing, to post-delivery, based on four core dimensions: safety policy formulation, whole-process risk management, safety capability verification, and dynamic iterative optimization. During the research and development and validation phase of autonomous driving systems, automakers must complete simulation testing, closed-track testing, and public-road testing in accordance with the specifications, ensuring that test resources and validation conditions meet the safety verification requirements before product launch. The standard also defines the safety boundaries for executing dynamic driving tasks in various extreme scenarios and provides normative guidance on the conditions and execution procedures for triggering the vehicle's minimal risk maneuver.

For L3 autonomous driving systems, the standard requires the installation of continuous driver takeover capability monitoring to prevent users from enabling high-level autonomous driving when they are unable to take over. In addition to providing product user manuals with the vehicle, automakers may also inform users through public channels such as official websites and in-vehicle central control terminals about the driving automation level, functional coverage scenarios, known functional limitations, control strategies in special scenarios, and user role transition rules in different states. The standard establishes a multi-faceted inspection and testing system encompassing enterprise safety capability verification, whole-process safety file validation, and product confirmation testing, to verify core dimensions such as the full-chain safety assurance capability of automakers and the completeness and robustness of product safety files, ensuring that high-level autonomous driving models that do not meet safety requirements cannot pass admission.