“I paid RMB 400,000 for this car and its driver assistance is worse than a domestic model costing a fraction of that.”
That is the experience reported by hundreds of European owners of Chinese cars. Soon after delivery, the assistance system began to misbehave, at times braking hard for obstacles that were not there. The manufacturer tried repeatedly to fix it but never produced a software patch that satisfied EU whole-vehicle type approval (WVTA). In the end it permanently disabled several assistance functions over the air and paid owners cash compensation.
Across China’s smart car exports, this kind of trouble is common. Liu Xiaobei, chief technology officer of Chelun Technology, surveyed the connectivity status of Chinese exported vehicles and found that around 70 per cent are offline. Of the roughly 30 per cent that are connected, only about one in ten owners renews the data subscription in the second year.
From assistance functions being switched off to cars that never connect after delivery, the test facing Chinese smart cars overseas has moved beyond the sale and into the years that follow it.
Does the car freeze on delivery day?
In China, a headline selling point of a smart car is that it keeps improving. After delivery, automakers push new functions over the air and fix software faults, so a car bought in year one is not the car driven in year three.
Overseas, that model runs into data-flow rules and local regulation. Zhang Xiaodong, deputy director of Geely Auto Group’s technical management centre, told an industry conference on export models that restricted cross-border data flow affects over-the-air updates for vehicles abroad.
Software updates also matter for how the system performs on real roads. Assistance software meets different road geometry, different road users and different driving habits every day, and the engineering team needs the vehicle’s actual behaviour to find the faults.
A connected car produces roughly 10TB of data a day, five to ten times the volume of a conventional combustion vehicle. That data covers road environment, driving behaviour and vehicle state, and it is the raw material for product analysis and algorithm work.
Cars abroad are also a source of real-road data. The longer they drive in a market, the more the automaker needs to understand local roads and local use. If that data cannot enter the development process, the post-delivery iteration loop loses its feedback.
Connectivity also gates services. Chinese automakers have folded online functions, software services and continuous upgrades into user operations, and some of those services only work while the car is connected. A permanently offline car cannot sustain them, and the automaker cannot build recurring revenue around them.
The difference is visible to owners. One reviewer noted that a Chinese-built electric car bought abroad has no CarPlay, so navigation runs through a third-party phone-mirroring app, and mirroring is restricted once the car is moving.
Assistance functions show the same pattern. The same system, sold into different countries, exposes different features, because local regulation, product certification and road conditions all constrain what can be enabled. The high-end capability advertised in China may be only partly available abroad, and the gap lands on brand perception.
Zhang Yanhua, president of the China Automotive Maintenance and Repair Association, argues that Chinese brands have already crossed the product-credibility threshold and that the next competitive frontier is aftersales, repair systems, insurance support and whole-lifecycle management.
This is a rules problem, not a capacity problem
For years the main export playbook was localisation: build the plant, open the channels, pull the supply chain in. A plant does not solve the offline problem.
Smart car exports also meet each country’s data rules. The EU General Data Protection Regulation sets strict conditions on processing and cross-border transfer of personal data, and vehicle operating data, in-cabin camera footage and captured street imagery can all involve personal information. Australia’s Privacy Act and several Southeast Asian regimes impose their own storage and transfer requirements.
South Korea is more complex. It restricts data leaving the country and also sets thresholds on call frequency and collection volume.
Those rules reshape corporate data pipelines. Chu Huabin, general manager of Qinda Rui (China) Information Technology, gives an example: an assistance engineering team road-testing in Southeast Asia can generate hundreds of gigabytes of street imagery, point-cloud data and decision logs in a day.
The team needs that data to analyse roads and debug algorithms, but faces and licence plates in the imagery are personal information. Raw data cannot leave the country, so it must be de-identified locally and plates must be tokenised.
Data processing therefore fragments across markets. Chu says some companies, weighing procedural complexity against violation risk, choose to keep data local, which makes it harder for domestic teams to obtain overseas data and slows algorithm iteration. Each new market means learning a new rulebook and building a new pipeline.
The friction extends to physical parts. Wen Hua, deputy general manager of Dongfeng Liuzhou Motor’s import and export company, says components that fail on overseas vehicles usually have to be shipped back for analysis, but some markets restrict imports of used parts, so the paperwork can delay the return and stretch the quality-improvement cycle.
Chinese automakers are not alone in this. Rivian’s published policy states that its service coverage is mainly the United States and Canada, and that connectivity and remote diagnostics outside those regions may be limited.
But connectivity is one of the competitive advantages Chinese cars have built in recent years, and overseas business keeps expanding. Europe deserves particular attention. Chinese brands reached 12.7 per cent of the EU market in the third quarter of 2026, overtaking Japanese brands for the first time, and a larger installed base means more vehicles, users and road data inside the bloc.
A plant is buildable. A research system is not
If overseas data cannot come home, automakers may have to deploy compute and engineering capacity locally, and that is expensive. Lu You, product and marketing general manager of NeueHCT, has costed it: a single overseas data centre at comparable scale runs into the hundreds of millions of renminbi, and another export executive puts the initial spend for re-deploying compute and models in one market at tens of millions.
Meanwhile Chinese automakers face compressed development cycles and falling vehicle prices, so they must hold research cost down while keeping iteration speed up. Rebuilding a research stack for every market is hard to recover from a single market’s volume.
Inconsistent data standards add friction. Hou Weiqiang, vice president of sales for LexisNexis Risk Solutions China, describes the mismatch: the same car carries a development code at the automaker, an industry model code in insurance, and another code at the parts supplier, with no reliable mapping between them.
That gap affects insurance pricing abroad. British drivers pay an average GBP 646 a year to insure a car, while Chinese-built cars exported to Britain average GBP 901, about 40 per cent more. Hou argues a major reason is that local underwriters do not hold enough information about Chinese imports.
Insurers need to know a vehicle’s specification to judge accident risk and repair cost. Without that, and without enough information to assess the risk of the technology, they price conservatively, and the smart features Chinese automakers add can end up scored as higher repair risk.
Chinese policy has begun to offer a defined route. In early 2026 eight ministries including the Ministry of Industry and Information Technology issued the 2026 edition of security guidelines for outbound automotive data, setting out three paths (security assessment, standard contract, personal information protection certification) and nine categories of exemption.
Important data covering security vulnerabilities, security incidents and the source code of over-the-air update packages is exempt from security assessment. The guidelines also scale requirements to the volume of personal information leaving the country, and companies declare data types and volumes rather than the data itself.
That gives firms an operating basis and reduces uncertainty, but each company still has to build the governance system, staff the function and coordinate research, operations and service teams across countries.
The second half of going global is the invisible road
Three directions are emerging. The first turns individual compliance problems into a shared public service. Chu proposes “cross-border data as a service”: a single entry point and end-to-end contracts that cut the cost of studying each market’s rules and building a pipeline. He says his team has already built eight categories of trusted automotive data space in Hainan and mapped 86 industry data scenarios.
The second is local infrastructure. Geely consolidated European engineering resources into a European technology company, one of whose jobs is meeting local data and cybersecurity requirements. XPeng’s Munich research centre focuses partly on localising assistance and cockpit development. Cloud providers are offering regional models: Alibaba Cloud proposes a regional centre plus country nodes, and GAC deployed its Middle East regional node on Alibaba Cloud and built a connected-car system in Saudi Arabia to localise storage and certification.
The third is a common data standard across the chain. Hou argues the industry needs a data hub and unified coding so automakers, insurers and suppliers can identify the same vehicle, with reliable mapping between development codes, sales names, insurance model codes and part numbers.
Sun Xiaohong framed the goal for the 15th Five-Year Plan period at the same conference: stand firm abroad, and bring it home. Chinese automakers need to hold their ground overseas and, within the rules, feed the experience and data back into the development system.
The competition used to be about screens, compute and assistance features. It now runs through data collection, compliance processing and cross-border collaboration. Cars can be shipped abroad quickly. The research and service system behind them takes years.
Editor’s note: This is an adapted translation of the original Gasgoo report. It has been trimmed and restructured for readability for an international business audience. The full original (in Chinese) is at https://auto.gasgoo.com/news/202611/6I70477018C108.shtml.