ZoZo Auto Research has released the "2026 Automotive Functional Safety and Safety of the Intended Functionality Research Report".
Multiple Mandatory Standards for China's Intelligent Vehicles Upgrade Functional Safety Requirements from Recommended to Mandatory Access Conditions
In 2026, China intensively issued and advanced multiple mandatory national standards for intelligent vehicles, comprehensively strengthening Functional Safety (FuSa) and Safety of the Intended Functionality (SOTIF) requirements, thereby establishing a clear safety baseline for intelligent vehicles.
Major Mandatory Automotive Standards Related to Functional Safety in China in 2026
Source: ZoZo Auto Research "2026 Automotive Functional Safety and Safety of the Intended Functionality Research Report"
For instance, in September 2025, the Ministry of Industry and Information Technology (MIIT) released the draft for comments on the "Safety Requirements for Combined Driver Assistance Systems of Intelligent Connected Vehicles," which is China's first mandatory national standard targeting Level 2 driver assistance systems. It specifies that the functional safety and SOTIF requirements for combined driver assistance systems shall comply with the applicable requirements of GB/T 34590 (all parts) and GB/T 43267. Appendix C and Appendix D are included to elaborate on the functional safety and SOTIF requirements for these systems. According to the standard, the combined driver assistance regulations upgrade functional safety requirements from recommended standards to mandatory access conditions for the first time. When developing Level 2, Level 2+, and above products, mainstream automakers have generally adopted ISO 26262 functional safety requirements as a standard configuration to meet current regulatory requirements and future compliance expectations.
Functional Safety and SOTIF Design Strategies for Level 3 Autonomous Driving Systems
In February 2026, the MIIT released the draft for comments on the "Safety Requirements for Autonomous Driving Systems of Intelligent Connected Vehicles," which will replace GB/T 44721-2024 "Intelligent Connected Vehicles - General Technical Requirements for Autonomous Driving Systems" to become the first mandatory national standard targeting Level 3 and Level 4 autonomous driving systems. The standard specifies the technical requirements, assurance requirements, and same type approval for the autonomous driving systems of intelligent connected vehicles, and describes corresponding methods such as assurance requirement inspections, safety case inspections, and validation tests. It also makes clear requirements for the functional safety goals related to hazards in Level 3/Level 4 autonomous driving systems.
Functional Safety Goal Requirements for Hazards Related to Level 3 ADS
Source: ZoZo Auto Research "2026 Automotive Functional Safety and Safety of the Intended Functionality Research Report"
In December 2025, China's first batch of Level 3 conditional intelligent driving vehicle models officially obtained access approval. The Arcfox Alpha S6 (Level 3 version) and Changan Deepal SL03 commenced on-road pilot programs in designated areas in Beijing and Chongqing, respectively, marking the beginning of the implementation of Level 3 conditional intelligent driving in China's passenger car market.
Relying on the "Tianshu Intelligence" technology system, Changan Automobile, in terms of process assurance for functional safety and SOTIF of Level 3 autonomous driving models, utilizes an ASIL D management and development process to build a seven-fold redundancy architecture encompassing braking, steering, power supply, communication, perception, control, and interaction, achieving a maximum system safety level of ASIL D. In terms of perception, it adopts a five-fold perception fusion matrix including high-resolution 4D imaging millimeter-wave radar, vision, and ultrasonic sensors. In actual road tests on the Inner Ring Expressway in Chongqing, the recognition distance for stationary obstacles reached 200 meters, a 40% improvement over Level 2 systems. More critically, when any single sensor fails, the system can initiate a backup plan within 0.3 seconds, a reaction speed 6 times faster than that of a human driver.
In terms of human-machine interaction safety, a driver takeover early warning mechanism is established: Considering the characteristic that drivers are prone to fatigue under congested traffic conditions, a progressive warning strategy is adopted, escalating step-by-step from mild prompts to strong alarms.
Changan Automobile's Full Lifecycle Functional Safety Strategy for Level 3 Autonomous Driving Models
Source: ZoZo Auto Research "2026 Automotive Functional Safety and Safety of the Intended Functionality Research Report"
In addition to Changan and BAIC, automakers such as Li Auto, BYD, XPeng, Xiaomi, and Voyah are also actively conducting road tests for Level 3 conditional intelligent driving models and advancing the on-road pilot processes for their Level 3 models.
In April 2025, targeting Level 3 intelligent driving, Dongfeng Voyah released its first Level 3 intelligent architecture—Tianyuan Intelligent Architecture, which integrates two core intelligent technology clusters: the Qingyun Level 3 Intelligent Safety Driving Platform and the Kunpeng Level 3 Advanced Intelligent Safety Driving System. At the functional safety design level, this architecture meets the requirements of the highest automotive functional safety level, ASIL D. At the hardware level, it achieves full-link backup design, adopting dual-backup designs from sensors and communication channels to computing chips and key execution components of the drive-by-wire chassis, ensuring that the vehicle can still maintain basic safe driving capabilities in the event of a single system failure. In terms of systematic active intelligent safety, the Kunpeng Level 3 intelligent driving achieves the leap from passive safety to active safety through global fusion of intelligent reasoning and autonomous learning.
Additionally, in terms of human-machine safety design, a three-level progressive warning mechanism has been created: when the system encounters special situations requiring manual takeover, the vehicle will issue three-level progressive warnings through interior light flashing, voice prompts, and seat vibration. If no one takes over for a long time, the vehicle will automatically and safely pull over. When encountering complex traffic environments with many pedestrians, the vehicle will remind people outside the vehicle to pay attention to safety through external speakers and lights.
Functional Safety Design Strategies for Steer-by-Wire Systems
China's new mandatory national standard "Motor Vehicle Steering Systems - Basic Requirements" will be implemented on July 1, 2026, fully replacing the current GB17675-2021 standard. To address new specialized technologies such as Steer-by-Wire (SBW) and Electric Power Steering (EPS), the new standard deletes the mandatory requirements for related mechanical connections, shifting the focus from mechanical structures to functional safety. Regarding the functional safety of steer-by-wire systems, the new standard emphasizes:
Mandatory requirement that the steering electronic control system must comply with functional safety international standards such as GB/T 34590 (all parts) (ISO 26262) and achieve the corresponding ASIL level (typically high levels such as Level D).
Strengthening redundancy capabilities: Clearly requiring that the steering system must possess redundant backup capabilities after a failure to ensure the vehicle can enter a safe state.
Refining failure responses: For full power steering systems, it details safety strategies, degradation processes, and alarm mechanisms under various failure scenarios such as power source failure, control signal transmission failure, and insufficient energy storage.
Mandatory Functional Safety Requirements for Steer-by-Wire Systems
Source: China Automotive Standardization Research Institute
The core focus of achieving functional safety compliance for steer-by-wire is: under the mandatory ASIL D level framework, through heterogeneous and fully redundant hardware and software design, coupled with millisecond-level fault diagnosis and processing mechanisms, and undergoing extremely rigorous fault injection and redundancy switching verification, it ultimately ensures that the system can still maintain controllable steering functions or safe parking when any single or even multiple reasonably foreseeable failures occur. As one of the hot technologies for the future development of intelligent vehicles, major suppliers and OEMs are actively deploying compliant steer-by-wire system products.
The steering system in ZF's Smart Chassis 2.0 completely eliminates the steering column. Through the coordinated operation of the steering wheel actuator, redundant steering front axle actuator, and the self-developed vehicle motion control software cubiX, it achieves full electrical signal transmission of steering commands. At the safety level, the fully redundant design meets the ASIL D safety standard. Coupled with the spatial advantages brought by the elimination of the universal joint intermediate shaft, it can effectively reduce the risk of leg injuries during a collision. Meanwhile, the system internally contains two sets of heterogeneous software and hardware as redundancy. When one system fails unexpectedly, the other system automatically and seamlessly connects, still allowing the vehicle to achieve complete steering functions.
In the intelligent chassis pre-research technology released by Xiaomi Auto, Xiaomi's 48V steer-by-wire system has no mechanical connection between the steering wheel and the wheels. The steering ratio can be continuously adjusted between 5:1 and 15:1, balancing the flexibility of low-speed turning and the stability of high-speed lane changes. It enables a steering-wheel-free cockpit layout, uses "wire" connections to achieve human-machine decoupling, and natively supports full autonomous driving. Its design meets the industry's highest functional safety level standard of ASIL D.
Exploration of AI Functional Safety and SOTIF Solutions Assisted by Automotive AI Safety Standards
Given the extensive application of AI large models and AI Agents in the automotive field, functions such as autonomous driving and intelligent cockpits are gradually becoming popularized, making the safety of automotive AI systems a focal point of attention. Issues such as the unexplainability, data dependency, and potential systemic failures of AI systems have made it urgent to address the safety problems of applying AI systems in automobiles.
In January 2025, ISO released the unified safety standard for AI system safety, ISO/PAS 8800:2024, aiming to regulate the application of AI technology in the automotive field, ensure its safety, reliability, and compatibility, and synergistically promote the development of AI system safety in conjunction with ISO 26262 automotive functional safety, ISO 21448, and information security.
In 2025, based on in-depth interpretation of standards and engineering practice, Hirain Technologies constructed a "three-in-one AI safety solution comprising safety process, safety framework, and safety platform", providing full-chain technical support for the safe development of intelligent vehicles. Targeting the characteristics of AI electrical architectures such as application-driven, layered decoupling, safety fusion, and innovative expansion, Hirain Technologies proposed a unified definition of the safety framework scheme, abstracting the implementation of software functional safety for AI systems into four key technologies: safety communication, safety isolation, safety monitoring, and safety execution. Through platform-based functional safety middleware and safety component technologies, the implementation difficulty is significantly reduced.
In August 2025, facing challenges such as AI model safety in highly autonomous driving (L3-L5) systems, Yaxi Shanghai, together with Saiweike, proposed the Distributed Weight Twinning (DWT) technology innovation solution to solve the difficult problem of autonomous driving AI redundancy development. DWT technology can prune a single E2E model into collaboratively operating "twin models", achieving safety redundancy through core technologies such as Weight Mirror Replication (WMR) asymmetric hardware deployment and neural bridging technology. Through DWT, a 40% cost reduction and over 99.99% fault coverage rate are achieved.
Problem-Solving Capabilities of Yaxi Shanghai's Distributed Weight Twinning (DWT) Technology Solution
Source: Yaxi Shanghai
[Contact Information] Mobile number is the same as WeChat number
Industry Research Department | Mr. Fu 15810027571
Mr. Zhao 18702148304
Data Service Department | Ms. Zhang 13716037793
Strategic Consulting Department | Ms. Han 15810133447
Promotion and Communication Department | Ms. Liao 13718845418
Mr. Du 13910162318