China has recalled 121 million vehicles cumulatively by the end of 2025, with 53% of those recalls triggered by regulatory defect investigations, according to Xiao Lingyun, director of the industrial products institute at the State Administration for Market Regulation's Defective Product Recall Technology Center. Speaking at the 2026 TEDA Auto Forum on September 19, Xiao said that with 370 million vehicles on Chinese roads by June 2026, roughly one in three car owners has experienced a recall.
The numbers sound alarming, but Xiao paired them with another data point: between 2019 and 2025, the fire rate per 10,000 new-energy vehicles fell by 68%. More recalls, fewer fires—the two trends together suggest that safety oversight has been effective over the past few years, he argued. Yet the real challenge is only beginning, as smart connected vehicles are creating problems that fall outside the scope of traditional recall rules.
One day, 8 million recalls
On August 21 this year, China saw nearly 8 million vehicles recalled in a single day, Xiao noted. Most of these were not traditional mechanical failures but issues tied to new technologies: software bugs, battery system problems, and flaws in intelligent driving functions.
The existing recall framework was built for the fossil-fuel era. The Product Quality Law defines what counts as a defect, and the regulations on recall of defective automobile products set out the process. That system has worked well for over a decade. But new questions keep emerging: battery-swap models are spreading, yet the rules only cover complete vehicle makers—who is responsible for swap batteries? Autonomous driving systems are increasingly concentrated in a few suppliers, and vehicle makers have limited visibility into those systems—can suppliers be held liable? Cybersecurity, data privacy, and information security issues, which did not exist in the traditional car era, now occur daily.
The recall regulations are being revised. Xiao said the new rules will bring motorcycles into scope and consider the regulatory needs of smart connected vehicles. But revisions take time, and the industry cannot wait. Last year, the market regulator and the Ministry of Industry and Information Technology jointly issued a document introducing requirements on accident reporting and sandbox regulation. Xiao called that document a milestone in smart-car safety oversight.
Traditional recall logic is reactive: an accident happens, consumers complain, regulators investigate, and a recall follows. That process has a built-in lag—by the time an accident occurs, people may already be hurt. Sandbox monitoring changes that. A sandbox is a dedicated extreme-stress testing environment outside real-world usage, where vehicles are pushed to their limits to uncover problems before accidents happen.
So far, about 200 vehicles have been tested in sandboxes and problems were found, Xiao said. Companies used those findings to improve their products proactively, preventing those flawed models from reaching the market and causing more accidents or injuries. Mainstream automakers' new intelligent models are gradually joining sandbox testing.
Behind this is what Xiao calls the "three-line theory." The sandbox is the high line, identifying risks before accidents; recalls are the bottom line, for when problems are already exposed; and in between is the daily accident reporting and monitoring. Together, the three lines form a safety oversight system for smart connected vehicles.
The technology center also introduced an accident reporting system last year, requiring companies to report intelligent driving accidents on time. Regulators aggregate and analyze the reports to build a database of dangerous scenarios. Before that system existed, the industry believed assisted driving problems were largely solved. But real-world data showed that vehicles still struggle to recognize and handle irregular or special objects. The reason is simple: lab tests are based on standard scenarios, while drivers face non-standard ones every day.
AI and standards fill the gap
The volume and growth rate of risks in smart connected vehicles far exceed what manual analysis can handle, Xiao said. His team is using large models and data feedback loops to boost regulatory efficiency. They are applying for two national standards: one for determining defects in assisted driving systems (to judge whether L2 and L3 production vehicles have design flaws), and another for accident reporting, standardizing the data format and content companies must submit.
More advanced work involves building AI-based product analysis tools. Regulators have already connected the accident reporting system to the dangerous scenario database, allowing the system to automatically generate new test scenarios based on historical accident data. Seven or eight companies are currently testing these auto-generated scenarios in sandboxes. Xiao said the goal is to enable quality regulators to quickly identify and detect risks in smart connected vehicles.
Standards are also being updated. The market regulator, in cooperation with CATL, released four national standards for power battery quality management last year. These standards were not written from scratch but distill the best management practices of leading companies into national norms. Next, they will be converted into international standards, giving China a voice in global rule-making.
Xiao said China's auto industry innovates fast, but many innovations fail—some consumers reject them, and safety issues can breed resistance. His judgment: in the end, only brands and models with both quality and safety will survive. That assessment is timely in a market where price wars are intense and companies are rushing to launch new cars and pile on features, often squeezing safety investment. The regulator's role is to hold the bottom line in such times.
Beyond domestic oversight, the defect product management center is building an overseas compliance platform. As Chinese car exports grow, data from overseas markets does not flow back, and companies are unfamiliar with warranty regulations and recall rules in target markets, leaving them vulnerable. The platform integrates regulatory requirements and data models from major overseas markets to help companies consider compliance at the design stage.
"Safety truly defines a car," Xiao said in closing. "Only safe products can survive in the market." The words sound simple, but in 2026, as intelligent driving accelerates, they carry weight. According to First Electric (D1EV), the content was originally published by Gasgoo.
