As the automotive industry continues to push the limits of R&D speed, regulatory authorities are drawing clearer safety baselines for new vehicle launches.
On October 10, the Ministry of Industry and Information Technology (MIIT) and three other departments released the "Notice on Further Strengthening the Management of Innovative Design, R&D Testing, and Verification for Automotive Products (Draft for Comments)" for public consultation. Fully hidden door handles are proposed to be banned, physical controls must be retained for critical functions such as lighting and wipers, and "zero-gravity" seats cannot be used while driving. More notably, reliability verification tests for new vehicles must cover no less than 30,000 kilometers, environmental adaptability verification cycles must be at least one year, and must cover "two winters and one summer".
Just a month ago at the 4th AI-Defined Vehicle Forum 2026, Wang Xianbin, Vice President of Gasgoo Research Institute, discussed the pressure that a 12-month car development cycle places on the supply chain: as the overall vehicle development cycle continues to shorten, the time left for some suppliers to complete R&D and verification may be only 3 to 4 months.
Car development can get faster and faster, but how much room is there left to compress safety-critical engineering verification?
Once-Hailed Innovations Begin to Face Re-evaluation
Over the past few years, a multitude of new designs have emerged in the NEV market.
Hidden door handles, half-yoke steering wheels, the elimination of physical buttons, zero-gravity seats, exterior light communication... these configurations have become important means for automakers to showcase technological appeal and build differentiation. However, as the scale of application expands, safety issues behind some of these designs have begun to surface.
Hidden door handles are a typical case in point.
Initially, this design primarily served aerodynamic and overall vehicle styling needs, later gradually becoming a signature configuration for NEVs. However, once subjected to low-temperature icing, electronic mechanism failures, or power loss after a collision, whether the doors can be opened in time is no longer just a matter of convenience.
On January 28, 2026, the State Administration for Market Regulation (SAMR) and the Standardization Administration of China (SAC) approved and released the "Safety Technical Requirements for Automobile Door Handles" (GB 48001—2026), which will take effect on January 1, 2027, setting clear requirements for the mechanical opening capability of door handles.
The current draft for comments further proposes banning fully hidden door handles and requires that door handles with innovative designs must feature mechanical redundancy and sufficient space for hand operation.
Similar changes are also occurring inside the vehicle.
As more and more automakers integrate functions like lighting and wipers into the central control screen, minimalist interiors have become a trend, but whether drivers can perform operations quickly and accurately in emergencies is easily overlooked.
Therefore, the draft proposes that for these driving-safety-related functions, even if touch operations are adopted, physical controls must be retained simultaneously. For features such as zero-gravity seats, rotating seats, and seats folded flat into a bed, it is required that they must not be used while the vehicle is in motion.
Regulation is not a blanket negation of automotive innovation, but rather a requirement for automakers to fully consider safety risks under abnormal operating conditions while pursuing differentiation in functions and experiences.
The fact that a design can be realized does not mean it is ready for mass production in vehicles. Nor does consumers' willingness to pay for new features mean they have been fully verified.
Building a Car in 12 Months: Where Does the Verification Time Come From?
Compared to the adjustments of a few popular configurations, the requirements for R&D testing proposed in this draft may have a more profound impact.
Over the past few years, shortening the development cycle has become a key competitive strategy for Chinese automakers. Traditional vehicle development often takes several years, but today some companies have compressed the cycle to about 18 months, and are even attempting to shorten it further to 12 months.
Behind this is the support of technological progress.
At the same forum, Meng Chao, Senior Director of the Intelligent Software Center of SAIC Passenger Vehicle and CTO of Z-ONE Technology, introduced that with the help of software architecture optimization, software-hardware decoupling, and OTA technologies, the overall vehicle development cycle can be shortened to 18 to 20 months, and the software development cycle can be further reduced to 6 to 8 months.
Platform-based development, reuse of mature components, digital simulation, and parallel development can indeed reduce repetitive work and improve efficiency.
However, some processes cannot have their time shortened by the same proportion.
The chassis needs to undergo repeated impacts under various road conditions, power batteries must endure high-temperature and extreme-cold tests, and the electrical and electronic (E/E) systems must also verify long-term operational stability in complex environments. Laboratories can simulate some operating conditions, but it is difficult to completely replace the complex environments of real roads.
For suppliers, the pressure may be even more direct.
A new vehicle involves multiple systems; from scheme determination and prototype development to overall vehicle integration, each phase has its own schedule. If the milestones for the overall vehicle project are continuously advanced, suppliers must complete development, testing, and rectification in a shorter timeframe.
Wang Xianbin stated bluntly: "Once all other time savings are exhausted and the milestones still cannot be met, the only option left is to cut into the verification process."
The current draft for comments is precisely aimed at holding this bottom line.
According to the proposed requirements, reliability verification tests for new vehicles must cover no less than 30,000 kilometers, environmental adaptability verification cycles must be at least one year, covering the "two winters and one summer" tests of extreme cold—extreme heat—extreme cold.
The document allows the use of off-season testing, environmental chamber simulation, and other methods, but must ensure that the verification effectiveness is no less than that of actual road testing. For complex environments that are difficult to equivalently substitute, real-vehicle testing must still be conducted.
More critically, if the relevant R&D verification is not fully completed, or if safety hazards discovered during testing have not been resolved, the vehicle model cannot apply for product market access.
These requirements are not only aimed at NEVs. Internal combustion engine (ICE) vehicles are also within the regulatory scope and must complete applicable reliability, durability, and environmental adaptability verification.
Of course, a one-year environmental verification cycle does not mean the new vehicle development cycle must exceed one year. Through reasonable arrangement of parallel testing and increasing the commonality of platforms and components, automakers still have room to improve efficiency.
The difference is that in the future, they can no longer omit verification work that should have been completed under the pretext of rushing to meet launch milestones.
AI Enters the Vehicle: Safety Verification Is Harder Than Ever
Compared to traditional mechanical structures, the challenges faced by smart vehicles are not just about having more and more functions, but that the way some functions operate is changing.
In the past, automotive software mostly executed tasks according to rules pre-set by engineers, and companies could conduct tests around input conditions and expected results. However, with end-to-end large models and AI agents entering vehicles, the situation has become complex.
Take advanced driver assistance systems (ADAS) as an example. For the same road scenario, subtle changes in weather, lighting, the behavior of other vehicles, and sensor inputs may cause the model to make different judgments. Even if the system performs well in a large number of routine tests, it does not mean it can definitely handle rare extreme situations correctly.
This brings a new challenge: when the system no longer makes decisions entirely according to pre-set rules, how should automakers prove it is safe enough?
Engineering practices from companies like FAW-Volkswagen show that intelligent driving cannot do without extensive real-vehicle testing. However, as the capabilities of AI models continue to expand, simply increasing test mileage may not necessarily cover all risks.
Some truly dangerous situations may precisely be the special scenarios that rarely occur in daily driving. Therefore, automakers must not only accumulate test mileage but also proactively seek out potential system failures through simulation, scenario replay, and real-vehicle verification.
The AI infrastructure solution previously shared by Meng Chao incorporates real-world scenario data, model training, evaluation verification, and safe deployment into a continuous iteration process.
This verification system is particularly important in the OTA era. Because a single software upgrade may change the vehicle's original control logic or system behavior, even if it passed previous tests, it is necessary to evaluate whether the new version introduces other risks.
The draft explicitly requires that combined driving assistance, autonomous driving, and smart cockpit systems must undergo real-vehicle testing, covering routine usage scenarios and extreme scenarios where safety hazards may exist. OTA upgrades involved also require sufficient testing and verification, and internal corporate approval records must be retained.
In addition, automakers must truthfully explain functional boundaries, must not exaggerate in promotions, and even more so, must not induce drivers to neglect their safe driving responsibilities.
For smart vehicles, the launch of a function does not mean the end of R&D verification. As software continues to update, safety assessments must also run through the entire vehicle usage process.
OEMs Can No Longer Directly Accept Suppliers' Verification Reports
If stricter testing requirements will affect the pace of automotive R&D, then the further clarification of responsibility attribution may change the cooperation model between OEMs and suppliers.
The draft proposes that for critical software and hardware such as advanced driver assistance, smart cockpits, power battery packs, motor controllers, and motors, OEMs should conduct incoming inspections and must not directly accept suppliers' verification reports.
At the same time, automakers should also completely preserve the original records, reports, and relevant supporting materials of R&D testing and verification, with a retention period of no less than 6 years.
This does not mean that suppliers' test results are valueless, but rather requires OEMs to bear their own quality management responsibilities.
In recent years, the division of labor in China's automotive industry chain has become increasingly deep. More and more Tier 1 suppliers are no longer just providing single components, but are participating in overall vehicle function definition, software development, and even system integration.
Especially in fields like intelligent driving and brake-by-wire, the safety performance of a system often depends on the synergy among different hardware, software, and control strategies.
The fact that a single component meets performance requirements does not mean it will function normally under all complex operating conditions after being installed in a vehicle.
Previously, discussions in the automotive industry surrounding product controversies such as braking systems have repeatedly touched upon similar issues: when a function malfunctions, how exactly should responsibilities be distinguished among component design, software control, and overall vehicle integration?
For increasingly complex automotive systems, it is already difficult to fully explain the overall vehicle's safety performance based solely on the test results of a single component.
Therefore, the new requirements do not ask OEMs to repeat all of the suppliers' tests, but emphasize that they must establish their own inspection and quality management capabilities, and cannot simply transfer overall vehicle safety responsibilities to suppliers.
This change will also raise the engineering threshold for some automakers who rely on suppliers' mature solutions to quickly integrate new vehicles.
In August this year, the MIIT and three other departments launched a one-year special action to improve production consistency and quality of automotive products, focusing on rectifying issues such as reliability, durability, and testing verification of new technologies. This draft for comments further shifts regulatory requirements forward to the R&D design and product market access stages.
For automakers and suppliers, stricter verification requirements will inevitably bring new costs.
Complex operating condition testing, long-term reliability data accumulation, and system-level verification all require investment in equipment, personnel, and capital. For companies with limited R&D resources that rely on quickly launching new models to seize market share, the pressure may be even more pronounced.
On the other hand, companies that have long valued engineering verification and possess a complete quality management system may also gain new competitive advantages.
In the future, automotive suppliers will compete not only on performance, price, and delivery speed, but also on reliability accumulation and the ability to support OEMs in completing system-level verification.
For OEMs, R&D efficiency cannot be measured solely by the development cycle. Even if a new vehicle is launched ahead of schedule, if it subsequently experiences frequent quality issues or even requires recalls, the time saved in the early stages may not have truly created value.
According to the proposed schedule in the draft, starting from January 1, 2027, newly declared vehicle models involving relevant requirements must supplement technical parameters and verification materials; relevant models that have already obtained announcements should complete supplementary reporting before July 1, 2027. Currently, the document is still in the public consultation stage, and the final regulations are yet to be determined.
However, the regulatory direction is already clear.
Over the past few years, the speed advantage established by China's automotive industry through supply chain synergy, technological innovation, and improved R&D efficiency does not need to be negated.
What truly needs to be guarded against is that when market competition turns speed into the most important metric, safety verification is also treated as a cost that can be continuously compressed.
Cars can iterate as quickly as consumer electronics, but they cannot be developed using the fault-tolerance logic of consumer electronics.
In the next stage, what China's automotive industry needs to prove is not just whether it can build a car in 12 months, but whether it can still complete sufficient verification within a shorter development cycle.
After all, car manufacturing can get faster and faster, but some time simply cannot be saved.