The "Notice on the National Recall of Automobiles and Consumer Products in 2025" (No. 2, 2026) issued by the State Administration for Market Regulation (SAMR) shows that in 2025, China implemented 190 automobile recalls involving 6.846 million vehicles. Throughout the year, recalls implemented via Over-The-Air (OTA) updates involved 1.756 million vehicles, accounting for 25.7% of the total number of recalls.
This means that automobile recalls are no longer limited to replacing parts; software upgrades have also become a crucial method for defect resolution.
However, the number of recalls does not directly indicate the quality of an automaker. When a car malfunctions, why do some only require repairs while others must be recalled? When issues involve software control or power batteries, how should enterprises assess risks and determine the recall scope?
01. Why Does a Car Malfunction Not Necessarily Require a Recall?
A car malfunction does not equate to a product defect in the context of a recall. According to Article 3 of the "Regulation on the Administration of Defective Automotive Product Recalls," a product defect refers to the unreasonable danger that endangers personal and property safety, or the widespread non-compliance with national or industry standards for ensuring personal and property safety in automobiles of the same batch, model, or category, caused by design, manufacturing, labeling, or other reasons.
Image source: Screenshot of the regulation
The key to this definition is not whether the vehicle has malfunctioned, but whether the issue involves product safety and meets the criteria for defect identification. For example, if a car experiences an increased braking distance, the cause could be a lack of necessary maintenance in the braking system, an anomaly in a specific component, or a design or manufacturing issue that poses a similar safety risk to vehicles of the same batch or type.
The first two scenarios require diagnosis and handling based on specific causes, but the possibility of a product defect cannot be ruled out merely because the malfunction occurred in a single vehicle. If an investigation confirms that the issue constitutes a product defect in the legal sense, the enterprise must further determine the scope of affected vehicles and implement a recall in accordance with the law.
It should be noted that a recall does not require an accident to have actually occurred, nor does it mean that every affected vehicle must have already malfunctioned. A product defect may only be exposed under specific temperatures, operating states, or usage conditions. The fact that some vehicles have not yet shown abnormalities does not prove that there is no risk.
Therefore, automakers need to comprehensively analyze information such as user complaints, after-sales maintenance records, fault data, component testing, and experimental verification to determine the failure mechanism, scope of impact, and safety consequences of the issue.
For intelligent vehicles, analysis must also be conducted in conjunction with software versions, electronic control strategies, and the operating status of related systems.
There is also an easily overlooked distinction here: a low probability of malfunction does not mean the risk can be ignored; however, a single occurrence of an issue cannot directly prove that the same car model has a defect requiring a recall. The final judgment must still be based on technical investigations and regulatory requirements.
02. Why Must Automakers Sometimes Issue a Recall After Discovering an Issue?
The core of the recall system is to require producers to take corresponding responsibility for their product defects, rather than waiting until an accident occurs. According to the "Regulation on the Administration of Defective Automotive Product Recalls," upon learning that an automotive product may have a defect, the producer shall immediately organize an investigation and analysis, and truthfully report the investigation results. If the investigation confirms the existence of a defect, the producer shall immediately cease the production, sale, and import of the defective automotive products and implement a recall.
Image source: Screenshot of the regulation
If the regulatory authority confirms the existence of a defect through investigation, it may also notify or order the producer to implement a recall in accordance with the law.
In 2025, there were 55 automobile recalls in China influenced by the supervision of the SAMR, involving 2.934 million vehicles, accounting for 42.9% of the total number of recalled vehicles for the year. This data indicates that defect investigations by regulatory authorities are an important mechanism for driving automobile recalls, but it cannot be inferred from this that the relevant enterprises previously refused to recall or deliberately concealed issues.
For automakers, after discovering an anomaly, the truly difficult task is to determine which vehicles are affected by the defect and what measures should be taken.
An issue may involve a specific production batch or span multiple models or configurations. Enterprises need to trace vehicle production records, component supply batches, software and hardware versions, and related verification results to determine the applicable scope of the issue. If the recall scope is not accurately defined, vehicles at risk may be omitted; if the scope is blindly expanded without technical basis, it will also increase the pressure on maintenance organization and resource allocation.
In November 2025, BYD Auto Industry Co., Ltd. and BYD Automobile Co., Ltd. respectively recalled some Qin PLUS DM-i vehicles, involving a total of 88,981 units. The SAMR announcement pointed out that the relevant vehicles had consistency issues in the power battery pack during production, which could lead to limited output power of the power battery, and in extreme cases, the inability to drive in pure electric mode, posing safety hazards. This recall was conducted under the circumstance that the SAMR initiated a defect investigation. The disposal plan did not simply end with an OTA upgrade. The enterprise first upgraded the vehicle-side detection software for free via OTA. When an anomaly in the power battery was detected, the fault light would be illuminated to prompt the user to return to the store, and then the dealer would replace the power battery pack for free.
This case illustrates that defect investigations not only need to identify exposed malfunctions but also analyze potential risks and design disposal plans based on the defect mechanism. Software can help vehicles identify anomalies, but if the root cause of the defect lies in the consistency issue of the power battery pack, the software upgrade itself cannot be considered as having eliminated the defect. A recall does not end with the issuance of an announcement either. Enterprises need to notify vehicle owners, organize repairs or replacements, and submit relevant reports according to the filed recall plan; regulatory authorities will also supervise the implementation of the recall and evaluate the effect of eliminating the defect. Even if the recall plan has been completed, if there are still defective vehicles that have not been recalled, the enterprise must continue to implement the recall.
03. Why Is a Recall Still Required If a Software Upgrade Can Solve the Problem?
The continuously enhancing software capabilities of intelligent vehicles have changed the manifestation of defects and the way enterprises handle issues. Some defects in traditional vehicles require replacing components or adjusting manufacturing processes.
For software control-related issues, enterprises may update control logic, correct program errors, or add fault detection functions via OTA. However, it must be clarified that OTA is a technical means, while a recall is a product defect disposal mechanism; the two cannot be confused.
If a software issue constitutes a product defect in the legal sense, the enterprise must still fulfill its recall obligation in accordance with the law.
The ability to upgrade remotely does not mean that the recall procedure can be bypassed; similarly, a vehicle undergoing an OTA upgrade does not necessarily mean that a product recall has occurred.
In 2025, China implemented 13 automobile recalls via OTA, involving 1.756 million vehicles, accounting for 25.7% of the total number of recalls for the year. During the same period, enterprises reported 1,904 OTA upgrades, involving 140 million vehicle-times. The first set of data statistics recalls implemented via OTA, while the latter statistics OTA upgrade activities reported by enterprises; the two cannot be regarded as the same type of event. However, looking at these two sets of data together, it can be seen that OTA has been widely integrated into the continuous maintenance process of automotive products, but routine function updates and safety defect rectifications still need to be distinguished.
The automobile defect clue data for the same year also provides an observation angle directly related to intelligent driving. Information released by the SAMR shows that in 2025, 31,000 reports of automobile product defect clues submitted by consumers were received, of which issues involving driving assistance systems accounted for 3.2%.
This set of data is worthy of attention, but its meaning needs to be accurately understood. Defect clues submitted by consumers are one of the sources for regulatory authorities to identify potential issues, but it does not mean that all these clues have been identified as product defects, nor can they be directly converted into the failure rate or accident rate of driving assistance systems. However, it at least indicates that driving assistance systems have entered the monitoring scope of automobile product defect clues, and related issues need to be further judged through investigation and technical analysis.
For intelligent driving systems, this distinction is actually very important. Suppose a certain driving assistance function has a control logic defect in a specific scenario. The enterprise needs to find out which car models, hardware configurations, and software versions are involved in the issue, verify whether the repair plan can cover the problem scenario, and evaluate whether the upgrade introduces new risks.
If the defect involves hardware, or if safety hazards cannot be eliminated solely through software upgrades, the enterprise needs to take other necessary measures. This also shows that the quality traceability system of automobile enterprises must be extended from traditional component batch management to software and hardware version and functional status management. Only by accurately identifying affected vehicles and verifying the effectiveness of rectification measures can enterprises possibly complete the closed loop from defect discovery to risk elimination.
04. What Exactly Can a Single Recall Reveal About an Automaker?
For enterprises, the most direct impact of a recall is cost. The enterprise needs to bear the costs of eliminating the defect and necessary transportation expenses, and must also invest R&D, quality, after-sales, and supply chain resources to complete fault analysis, component replacement, software development, user notification, and subsequent verification. If a large number of vehicles are affected, maintenance capabilities, component supply, and service outlet coordination will all face pressure.
A recall may also affect users' trust in the brand, but this impact is not inevitable, and an enterprise's market performance cannot be inferred solely based on the number of recalls. Sometimes, how an enterprise discovers issues, assesses risks, and whether it has taken effective measures can instead demonstrate the enterprise's sense of responsibility.
In 2025, automobile recalls in China implemented due to design reasons involved 4.215 million vehicles, accounting for 61.6% of the total number of recalls for the year; recalls due to manufacturing reasons involved 2.502 million vehicles, accounting for 36.5%. The number of vehicles involved in design reasons is significantly higher, but this cannot directly prove that design defects are more severe than manufacturing defects. The scale of recalls is also affected by factors such as the scope of impact of a single defect, model sales, and investigation results. This set of data cannot be used to simply rank the R&D capabilities or quality levels of automakers.
Zhijia Zuiqianyan believes that a recall tests at least three aspects of an enterprise's capabilities.
The first is defect identification capability. Whether an enterprise can promptly discover anomalies and find potential common issues through complaints, after-sales, and vehicle operation data determines whether risks can be identified as early as possible.
The second is issue traceability and rectification capability. Whether an enterprise can accurately determine the affected vehicles and software and hardware versions, find the root cause, and design effective measures determines whether the recall scope is reasonable and whether the rectification can truly eliminate the risk. Then there is the quality closed-loop capability.
After the recall is completed, whether the enterprise feeds the issue back into product design, manufacturing processes, software development, and verification processes, and whether it can prevent the same issue from recurring, determines whether this recall can be transformed into long-term quality improvement.
Therefore, a high number of recalls does not mean poor quality management; a low number of recalls cannot directly prove better product quality either. Proactive recalls reflect the enterprise's attitude towards assuming product safety responsibilities, but the severity of the defect itself cannot be ignored just because the enterprise is proactive.
For the automotive industry, a recall is not merely an after-sales cost, nor is it a synonym for a public relations crisis. It is a set of mechanisms for enterprises to legally identify risks, eliminate defects, and accept supervision when safety issues are discovered after products are launched. The more complex automotive technology becomes, the less quality management can rely solely on pre-delivery inspections.
After vehicles enter the real usage environment, enterprises still need to continuously collect issue clues, conduct technical analysis, and take timely measures when defects are confirmed. What a single recall truly tests is not just whether the enterprise can fix the cars, but whether it can identify risks, find the root cause, eliminate defects, and transform the issues into improvements for the next round of product development and verification.