Gasgoo Munich-On Sept. 19 at the 2026 TEDA Forum, Xiao Lingyun took the stage with a stark set of numbers. As head of the industrial products division at the Defective Product Recall Management Center under the State Administration for Market Regulation, he noted a significant figure. By the end of 2025, China had logged 121 million vehicle recalls. More than half—53%—were triggered by regulatory defect investigations. By June 2026, with 370 million vehicles on the road, roughly one in three had been subject to a recall.
Image source: 2026 TEDA Forum
That figure alone sounds significant. But Xiao quickly followed with another data point: between 2019 and 2025, the fire rate for new energy vehicles dropped 68% per 10,000 units.
More recalls, fewer fires. Taken together, these figures suggest that safety measures over the past few years have been effective.
Yet the real challenge is just beginning. New issues arising from intelligent connected vehicles are already exceeding the scope of traditional recall systems.
Nearly 8 Million Recalled in a Single Day: The Cost of New Tech
Xiao pointed to a recent case. On Aug. 21 this year, the Chinese market experienced nearly 8 million vehicle recalls in a single day. Most were not due to traditional mechanical failures. Instead, they were linked to new technologies. These included software glitches, battery system flaws, and problems with intelligent driving features.
The traditional recall framework was built for the internal combustion engine era. The Product Quality Law defines "defects," while regulations on defective auto product recalls mandate the process. That system worked well for over a decade. Now, however, new issues are surfacing constantly.
Battery swapping is becoming more common, but regulations only target automakers—so who is accountable for the safety of swapped batteries? Autonomous driving systems are increasingly concentrated among a handful of suppliers. This leaves automakers with limited insight into system internals. Should suppliers bear liability? Cybersecurity, data privacy, and safety—issues virtually non-existent in the traditional auto age—are now daily occurrences.
Regulations are under revision. Xiao revealed that the new recall rules will include motorcycles and address the oversight needs of intelligent connected vehicles. But the legislative cycle is long, and the industry cannot afford to wait. Last year, the State Administration for Market Regulation and the Ministry of Industry and Information Technology issued a joint document. It outlined requirements for accident reporting and "sandbox" oversight. Xiao called this a milestone for smart vehicle safety. Traditional recall logic follows a linear path. An accident occurs, consumers complain, regulators investigate, and automakers recall. That process has a built-in lag—by the time an accident happens, people may already be hurt.
Sandbox oversight changes that dynamic. A "sandbox" is a controlled, high-stress testing environment separate from real-world driving. Vehicles are subjected to extreme conditions to push them to their limits. This allows regulators to proactively identify issues before accidents occur on the road.
According to Xiao, problems have already been identified in roughly 200 vehicles through sandbox testing. Companies have used these findings to make improvements. This prevents flawed models from reaching the market and causing further harm. Major automakers are gradually enrolling their latest smart models in the program.
This approach is backed by a "three-line theory." Xiao explained that the sandbox represents the high line—proactively spotting risks before accidents. Recalls act as the bottom line—mandatory action once problems surface. In between lies the middle line: daily accident reporting and monitoring. Together, these three form the safety oversight framework for intelligent connected vehicles.
The Defective Product Management Center also launched an accident reporting system last year. It mandates that automakers report smart driving incidents promptly. Regulators then aggregate and analyze this data to build a database of hazardous scenarios.
Before the reporting system was established, the industry believed issues with driver-assistance systems were largely resolved. But actual data revealed that vehicles still struggle to recognize and handle irregular objects. The reason is straightforward. Lab tests rely on standard scenarios, yet drivers face unpredictable, non-standard conditions every day.
Using AI and Data Loops to Keep Pace with Tech Iteration
The sheer volume and speed of emerging risks in smart vehicles far exceed the processing capacity of manual analysis used in the traditional auto era. Xiao's team is using large models and data loops to boost regulatory efficiency.
They are currently applying for two national standards. One focuses on defect identification for driver-assistance systems. It determines whether Level 2 and Level 3 production vehicles have design flaws. The other standardizes accident reporting, regulating the format and content of data submitted by automakers.
At the forefront, the team is using AI to build product analysis tools. Regulators have integrated data interfaces between the accident reporting system and the hazard scenario database. This enables the system to automatically generate new test scenarios based on historical crash data.
Approximately seven or eight automakers are already testing these auto-generated scenarios within the sandbox. The goal, Xiao said, is to enable quality regulators to rapidly identify risks associated with intelligent connected vehicles.
Standards are also evolving. The market regulator collaborated with CATL to release four national standards for power battery quality management last year. These were not created in isolation. They condensed the best practices of leading companies into national standards. The next step is converting them into international standards to secure a voice in global rulemaking.
Innovation in China's auto industry is moving fast, Xiao noted, but many attempts have failed. Some innovations never gained consumer acceptance. Others even triggered resistance due to safety concerns. He concluded that only brands and models that excel in both quality and safety will survive the market's inevitable consolidation.
That assessment is particularly relevant given the current market climate. Since 2026, a fierce price war has forced many automakers to focus on launching new models. They are adding features, often at the expense of safety investment. The regulator's role is to hold the safety line when the market gets this heated.
Beyond domestic oversight, the Defective Product Management Center is building an export compliance platform. Chinese vehicle exports are surging, but data from overseas does not always return. Companies often lack familiarity with local warranty and recall rules, putting them at a disadvantage. This platform consolidates regulatory requirements and data models from major overseas markets. It helps automakers factor compliance into the design phase.
Xiao concluded with a simple truth: safety is what truly defines a car. Only safe products can endure in the market. That may sound like a cliché, but in 2026, as the industry races ahead with autonomous driving, it carries significant weight.
