Woofun AI reports that the Ministry of Industry and Information Technology (MIIT) has officially promulgated China's first mandatory national standard for L3 and L4 advanced autonomous driving, titled 'Safety Requirements for Autonomous Driving Systems in Intelligent Connected Vehicles.' This regulatory instrument is scheduled to take full effect on July 1, 2027, mandating that all newly applied vehicle models equipped with L3 or L4 systems must strictly adhere to these requirements, while existing models are granted a one-year transition period to implement necessary adjustments. This standard represents a definitive upgrade from the previous recommended national standard GB/T 44721-2024, effectively terminating a prolonged era of unregulated development characterized by the absence of mandatory guidelines, ambiguous human-machine responsibility boundaries, and disjointed regional pilot programs. It establishes a unified safety framework covering the entire lifecycle of L3 conditional autonomous driving and L4 highly autonomous driving, serving as the critical mechanism linking technology research and development, commercial operations, and industry regulation.

Unlike prior advisory documents that offered only technical guidance, this national standard possesses legal binding force and applies comprehensively to passenger vehicles, commercial vehicles, and consumer-grade autonomous driving units. The regulation defines five core systems: system operation boundaries, human-machine division of labor, emergency response protocols for system failures, enterprise safety management structures, and standardized testing and inspection procedures.

Furthermore, it incorporates accompanying mechanisms for safety records, data retention, and full-process verification. For the industry, this document transcends the role of a mere technical testing specification; it functions as a top-level design that reshapes the competitive logic of the sector, removes barriers to commercialization, and significantly improves the supporting legal and regulatory infrastructure.

Woofun AI data shows that the standard's implementation will fundamentally alter the cost structure and entry barriers for all market participants.

Over the past five years, China's intelligent connected vehicle industry has achieved leapfrog development, with L3 models gradually obtaining road operation permissions and L4 systems being piloted regularly in diverse scenarios including ports, mining areas, industrial parks, freight logistics, and urban Robotaxi services. More than a hundred cities across the nation have granted qualifications for autonomous driving road tests, driving continuous market expansion.

However, the industry's long-term reliance on advisory standards and local pilot management methods has precipitated three critical systemic problems. First, widespread marketing confusion has emerged, where many automobile companies have misrepresented L2 assisted driving as 'quasi-L3' or 'fully autonomous driving,' deliberately obscuring SAE classification criteria. Some models lack sufficient hardware redundancy and proper takeover logic, leading to consumer deception through aggressive marketing rhetoric. Second, there has been no unified basis for determining responsibilities in human-machine accidents. Incidents involving systems operating outside designated ranges, drivers failing to take over in L3 scenarios, or emergency handling failures in L4 operational vehicles occur frequently. The absence of national standards defining responsibility boundaries has made resolving these accidents and processing insurance claims fraught with difficulties. Third, pilot rules vary significantly across different regions, with substantial differences in requirements for road autonomy, safety officer allocation, and test distances between first-tier cities and third- and fourth-tier cities. This fragmentation results in high cross-regional operating costs for enterprises and hinders the establishment of a unified, scalable model nationwide.

The global context further underscores the urgency of this regulation, as frequent autonomous driving safety incidents worldwide and tightening regulations on Robotaxi services in many countries demonstrate that the lack of strict safety measures poses serious risks to road traffic safety. At this critical juncture, as the industry transitions from pilot projects to mass production, the introduction of a mandatory unified safety standard is essential to ensure the safety of road users and regulate healthy industry development. Currently, Europe, America, Japan, and South Korea have all introduced mandatory regulations for advanced autonomous driving, while the United Nations is promoting global unified technical regulations. The competition among national standard systems is intensifying. In formulating this national standard, China has fully aligned with international standards such as ISO and SAE, while simultaneously accounting for complex road conditions and diverse transportation environments in urban and rural areas to develop a safety indicator system suitable for local conditions. Upon implementation, China will possess a complete set of independent rules for L3/L4 autonomous driving access, testing, and safety management, providing a basis for domestic automobile companies to operate overseas and enabling China to share its autonomous driving governance experience globally.

The state has continuously introduced top-level policies for intelligent connected vehicles, gradually establishing a four-tier governance framework consisting of 'top-level planning, local pilots, mandatory national standards, and supporting regulations.' This mandatory national standard fills the institutional gap in the access phase for mass-produced autonomous driving vehicles, connecting upstream component and vehicle-grade standards with downstream regulations related to road traffic, insurance, and traffic liability determination. This ensures comprehensive supervision throughout the entire chain from research and development, production, and sales to road operation. The standard consists of four core sections: basic safety requirements for autonomous driving systems, human-machine interaction and takeover management rules, minimum risk strategies in case of system failures (MRM), and enterprise safety assurance and inspection mechanisms for the entire lifecycle. All provisions include quantifiable and measurable indicators, eliminating vague or flexible wording.

The first core requirement involves the clear definition of the Design Operating Domain (ODD) and the establishment of system capability boundaries. The ODD serves as the foundation for distinguishing autonomous driving capabilities and defining safety responsibilities. The standard mandates that automobile companies publicly disclose all applicable boundaries for L3 and L4 systems, covering six dimensions: road type, speed range, weather visibility, lighting conditions, traffic flow density, and road surface condition. The system must continuously determine whether it is within the legal ODD range during operation. When a vehicle is about to leave the preset ODD or encounters conditions beyond the system's capabilities, such as low visibility due to rain or fog or icy roads, a graded warning must be triggered immediately, and the system must not continue to operate autonomously when incapable. Specific quantitative indicators have been set for L3 functions in high-speed scenarios: the minimum forward detection distance under a cruising speed of 120 km/h should not be less than 130 meters, and the lateral detection range should cover at least 9 meters on either side of the vehicle. In complex environments such as light rain or at night without street lights, the detection range should not decrease by more than 20%. These requirements promote the implementation of multi-sensor fusion solutions and help avoid safety hazards in adverse conditions caused by single visual perception.

The second core requirement dictates the hierarchical design of the human-machine interaction system and rigid constraints for L3 takeover rules. The standard establishes two completely independent human-machine interaction logics for L3 and L4 systems, with the primary distinction being whether human intervention is required as a backup driver. For L3 conditional autonomous driving, the driver remains the legally designated responsible party. The standard requires the installation of a multi-dimensional driver monitoring system (DMS), and the ability to take over cannot be determined solely based on steering wheel torque. At least two independent indicators, such as seat occupancy status, seatbelt wearing, eye gaze direction, and head posture, must also be checked. If the driver leaves the seat, becomes distracted, closes their eyes, or does not wear a seatbelt, an immediate audible and visual warning will be issued. If the driver fails to regain effective control within 10 seconds, the warning will escalate to seatbelt pre-tensioning and seat vibration alerts. The system reserves at least 10 seconds for a complete takeover process to avoid risks associated with forced takeover in short periods. In terms of responsibility allocation, if an accident occurs while the system is within the ODD range and no takeover request has been made, the automobile company shall bear the primary responsibility; if the driver fails to manually take over within the specified time after the system issues a takeover alert, the driver shall bear the corresponding responsibility. This clearly resolves the long-standing issue of determining responsibilities in human-machine accidents.

For L4 highly autonomous driving, the system independently handles all dynamic driving tasks. The core rule for L4 is that there is no need for human intervention inside the vehicle; even in vehicles equipped with steering wheels, passengers do not have a legal obligation to take over. Robobuses without steering controls, unmanned vehicles in mining areas, and Robotaxi services can operate in compliance with these regulations. The standard specifies that remote monitoring can only be used for post-event assistance and cannot be relied on to handle sudden road risks manually. All system failures, obstacles, and extreme conditions must be independently managed by the on-board system. In closed scenarios such as scenic areas, ports, and mining areas, L4 vehicles may be exempt from installing driver monitoring devices, but they must be equipped with complete automatic failure handling modules.

The third core requirement focuses on Minimum Risk Strategies (MRM) as a safety baseline for all scenarios. MRM constitutes the core safety provisions of this standard, establishing hierarchical response procedures for three types of risk scenarios: system failures, situations where the system operates outside the ODD range, and cases where the driver refuses to take over. Operations such as sudden hard braking and emergency lane changes, which can easily lead to secondary accidents, are strictly prohibited. During the warning reduction phase, the maximum driving speed is restricted, automatic lane changing and overtaking are prohibited, and the system is continuously prompted visually and audibly to reduce its functionality. In the emergency response phase, the vehicle decelerates smoothly and linearly, the hazard warning lights are activated, and it gradually moves to the emergency lane or stops by the roadside while ensuring the safety of surrounding vehicles. In the parking and locking phase, once the vehicle comes to a complete stop, the autonomous driving function is locked, and the driver must restart the power system to re-activate the advanced autonomous driving features, thereby avoiding the risk of repeated system startups and shutdowns in case of failures.

The fourth core requirement mandates safety records and enterprise safety management throughout the lifecycle. This standard introduces the internationally recognized Safety Case system for the first time. All models applying for L3/L4 qualifications must submit a complete and structured safety record, including a risk analysis report, reliability verification data for perception and decision-making hardware, simulation and real-road test records over tens of thousands of kilometers, human-machine interaction logic designs, failure emergency plans, and OTA upgrade management procedures. Testing institutions will verify each item individually, and models with incomplete records or data will be directly rejected from listing. Simultaneously, automobile companies are required to establish a comprehensive safety management system covering research and development, manufacturing, after-sales service, and remote upgrades. They must install dedicated data recorders for autonomous driving to accurately record data on system activation, fault reports, takeover requests, and vehicle collisions. The data must be retained for at least three years for accident tracing and regulatory inspections. Strict control mechanisms have been established for OTA upgrades; major algorithm updates must undergo comprehensive safety verification before being pushed to users, and automobile companies are prohibited from mass-producing unstable autonomous driving programs without testing them first.

To ensure the implementation of the standard, a three-layer closed-loop inspection system has been established. The first layer involves verifying the enterprise's safety management system, testing facilities, and simulation testing capabilities. The second layer involves thoroughly checking the enterprise's safety records and risk assessment materials. The third layer involves conducting performance verification under various conditions through simulation tests, closed-road tests, and real-road tests. Only when all three layers of verification are passed can a model obtain L3/L4 type approval qualifications, thereby raising the industry's entry barriers from the source. This national standard continues the overall regulatory approach of 'starting with closed scenarios and gradually expanding to open ones' for China's autonomous driving industry. It implements differentiated management based on different levels of automation and different operating scenarios, avoiding one-size-fits-all approaches and balancing safety requirements with innovation space.

In terms of automation levels, L3 is positioned as a transitional form for mass production in the consumer segment and is suitable for use on closed highways, with regulatory focus on human-machine interaction and driver takeover supervision. L4 is aimed at specialized operational vehicles, covering closed industrial parks, freight logistics, and urban mobility, with regulatory focus on the system's ability to independently handle risks, redundant hardware configuration, and safe operation in scenarios without safety officers. In terms of application scenarios, the standard relaxes some human-machine interaction requirements for closed/semi-closed scenarios such as ports, mining areas, industrial parks, and scenic area sightseeing vehicles, allowing vehicles without steering wheels or DMS monitoring devices to operate in compliance with regulations. For high-speed freight logistics scenarios, emphasis is placed on long-distance perception, formation driving, and braking redundancy indicators. For urban open-road Robotaxi and bus services, the most stringent requirements for perception redundancy, emergency response in extreme weather conditions, and data retention are applied. For household passenger vehicles, L3 is currently limited to use on high-speed roads and is not yet supported for widespread conditional autonomous driving in urban areas. The underlying logic of this differentiated regulation is to match the risk levels of different scenarios, prioritizing commercialization in controllable risk environments while imposing higher safety requirements on complex urban scenarios.

The implementation date of this standard in July 2027 will mark a turning point for the industry, reshaping development logic across vehicle manufacturers, upstream components, mobility services, and insurance and legal support. In the short term, it will increase compliance costs, but in the long run, it will accelerate industry consolidation and promote high-quality, large-scale development. The cost gap between leading automobile companies and smaller and medium-sized enterprises will widen. To meet mandatory requirements for perception redundancy, DMS monitoring, safety records, and simulation testing facilities, enterprises need to increase investment in research and development, testing, and hardware procurement. Smaller and medium-sized enterprises lacking technical reserves and financial strength will be unable to meet these requirements and will gradually withdraw from the L3/L4 market, leading to accelerated industry restructuring. Marketing strategies relying on simplifying hardware and exaggerating autonomous driving capabilities will no longer be effective. All promotional claims must correspond one-to-one with national standard testing indicators, and market promotion and supervision will be tightened accordingly.

In the long term, the unified national standard will reduce cross-regional expansion costs for leading automobile companies, eliminating the need to adapt to local differential pilot rules. Unified listing and operation across the country will become possible. Enterprises will shift their research and development focus from pursuing extreme functional displays to ensuring system safety redundancy, refining algorithms for extreme operating conditions, and managing risks throughout the entire lifecycle. Industry competition will shift from a 'parameter race' to a comprehensive comparison of safety, reliability, and commercial operation capabilities. The safety record and data retention mechanisms will help automobile companies establish complete failure iteration databases, accelerating the efficiency of data closed-loop iteration. For the upstream component industry, there will be increased demand for knowledge, hash rate, and monitoring hardware. The mandatory performance indicators for perception systems have directly established multi-sensor fusion as the mainstream technical approach. Single pure visual perception solutions will struggle to pass testing requirements in rainy, foggy, or dark conditions. Market demand for 77GHz millimeter-wave radars, lidars, high-specification cameras, and vehicle-grade hash rate chips will continue to grow. The requirement that the lane-level lateral error of positioning modules be ≤0.5 meters will drive the growth of industries related to high-precision maps and IMU inertial navigation. As the industry standardizes component testing indicators, component manufacturers will no longer need to customize adaptation solutions for different automobile companies, reducing hardware costs through mass production. After 2027, the prices of high-end autonomous driving perception hardware are expected to decline, laying a cost foundation for widespread adoption in the medium to long term. Supporting safety hardware such as data recorders and on-board storage will become standard equipment for L3/L4 vehicles, opening up new niche markets.

For operational enterprises such as Robotaxi providers, autonomous driving trucks, and unmanned buses, this mandatory national standard provides a legal basis for operating without safety officers on a regular basis. Previously, the lack of a unified standard made the approval process cumbersome. After implementation, L4 operational vehicles meeting Design Operating Domain, Minimum Risk Strategies, and redundant perception requirements will be able to simplify the road approval process, accelerating commercial replication in ports, mining areas, and high-speed freight logistics routes. The clear definition of responsibilities in case of accidents solves the biggest operational challenges. In the past, ambiguity in determining responsibilities led to difficulties in insurance underwriting and high premiums. Now that the standard clarifies responsibilities, insurance companies can accurately assess risks, and specialized autonomous driving insurance products can be quickly introduced, significantly reducing costs associated with accident claims and opening a key path to commercial profitability.

As the technical foundation for these regulations, the national standard will promote the revision of the 'Road Traffic Safety Law' to clarify the legal entities of autonomous driving vehicles. Judicial authorities can use these standard provisions to determine primary and secondary responsibilities in accidents, resulting in standardized judicial precedents. The insurance industry will rely on these risk indicators to establish a tiered and scenario-specific premium system, offering differentiated coverage for L3 household vehicles and L4 operational vehicles. The data retention mechanism will provide a complete and objective basis for judicial investigations of traffic accidents, significantly reducing the time required to resolve disputes. With only a little over a year left before official implementation, some automobile companies have not yet established complete simulation testing platforms or safety record management systems, and their redundant perception hardware design does not meet requirements. To address this, the standard sets a phased transition period: new models applying for certification starting from July 2027 must comply immediately, while existing models with certification have one year to make necessary adjustments. Enterprises can share resources through third-party testing institutions and joint simulation testing platforms to reduce individual investment costs.

The standard requires testing under a wide range of extreme conditions, including heavy rain, dense fog, sudden changes in tunnel lighting, and pedestrians crossing roads unexpectedly. The cost of conducting real-road tests under these conditions is extremely high. The industry can rely on national-level intelligent connected vehicle testing demonstration zones to jointly build and share extreme weather simulation facilities and simulation databases covering millions of kilometers. Industry associations can lead the establishment of joint testing platforms to share testing costs and accelerate algorithm iteration. In terms of supporting infrastructure, L3 high-speed autonomous driving and L4 urban operations rely on clear road signs and vehicle-road coordination devices.

However, some roads in China have outdated road signs and lack vehicle networking infrastructure, restricting the expansion of the system's design and operation range. In the future, transportation and MIIT departments will work together to upgrade road infrastructure, ensuring coordinated development of vehicles, roads, and standards.

The implementation of this mandatory national standard officially marks the end of the pilot phase for autonomous driving in China and the beginning of a new stage of large-scale development characterized by 'safety first, compliance first, and phased implementation in different scenarios.' The development of the industry will show three clear trends. First, implementation will follow a phased approach, starting with closed-road operations, then moving to high-speed commercial vehicles, followed by urban mobility services, and finally achieving widespread L4 autonomous driving in urban areas. L4 operations in ports and mining areas already have a solid legal foundation and will see rapid expansion after 2027; L4 autonomous driving trucks in freight logistics will become widespread within 3 to 5 years; urban Robotaxi and unmanned bus services will gradually achieve operation without safety officers across the city within about 5 years; however, L4 autonomous driving for general consumers in urban areas will be limited due to environmental complexity and regulatory constraints, and is not expected to be widely available until around 2035. Second, the industry will form a positive cycle of 'policy support, technological innovation, and commercial feedback.' The standard defines clear compliance boundaries, and policies will gradually lift restrictions on specific scenarios; enterprises will produce vehicles on a large scale using these standards, accumulate real-road data to continuously optimize algorithms; and commercial operations will generate stable cash flows, supporting further hardware research and development and safety technology improvements. Third, China's national standard system will continue to be exported globally, promoting the globalization of the autonomous driving industry. With these comprehensive mandatory safety standards, domestic automobile companies, autonomous driving solution providers, and component manufacturers will have a solid legal foundation for operating overseas. The mutual recognition of Chinese and foreign standards will continue to advance, and Chinese autonomous driving technologies, vehicle products, and operation models will enter the global market. For all participants in this industry, this standard is the fundamental basis for the widespread commercialization of autonomous driving. Vehicle manufacturers, component suppliers, and service providers must seize the one-year transition period to address gaps in safety technology capabilities and improve safety management systems throughout the entire lifecycle. Regulatory agencies, insurance companies, and transportation departments must also work together to update laws, regulations, financial products, and road infrastructure in a coordinated manner. Under the guidance of these unified standards, L3 and L4 autonomous driving will transition from a fragmented experimental phase to a regulated, scalable, and commercially viable reality, fundamentally reshaping the global automotive landscape.