China’s New Self-Driving Rules Raise the Bar for Tesla’s FSD Ambitions

Approved on July 30, 2026, GB 44721-2026 applies to passenger and commercial vehicles equipped with Level 3 conditional automation or Level 4 high automation, excluding automated-parking systems. For Level 3 vehicles, the system must monitor whether the user remains capable of taking control and issue a sufficiently early intervention request when it approaches the boundary of its operating conditions. The new framework includes a takeover period of at least 10 seconds in specified foreseeable situations. If the driver does not respond, the vehicle must execute a minimum-risk manoeuvre instead of simply returning control without protection. At Level 4, the system must handle the driving task and reach a safe state without relying on a human fallback. Manufacturers must demonstrate safety governance across product development, production and post-deployment operations, covering risk identification, safety assurance and continuous improvement.

The standard strengthens the technical basis for determining responsibility when an automated vehicle crashes, although it does not by itself resolve every civil-liability case. Data recorded by the vehicle can help establish whether the automated system was active, whether a takeover request was issued correctly and whether the driver responded. This is a major change from today’s Level 2 environment, in which the human driver remains responsible for continuously monitoring the road. China has separately published GB 47955-2026 for Level 2 combined driver-assistance systems, effective January 1, 2027. According to the Ministry of Industry and Information Technology, around 70% of new passenger vehicles sold in China in 2026 have some form of combined driver assistance, while more than 30% have navigation-assisted driving. The government is therefore building parallel regulatory tracks for existing assistance systems and genuinely automated vehicles.

Tesla’s immediate challenge is that FSD, despite its name, remains a supervised Level 2+ system. It can assist with steering, acceleration, braking and navigation, but the driver must remain attentive and ready to intervene. This means Tesla could prioritize a broader Level 2 rollout in China without waiting for Level 3 certification. However, calling the product “Full Self-Driving” creates a regulatory and communication problem in a market where authorities have restricted automakers from using terms such as “autonomous driving” to advertise systems that still require supervision. China also requires approval for significant over-the-air updates to driving-related software. Consequently, Tesla cannot rely on the same rapid software-release model it uses in North America; major improvements may require additional testing, documentation and regulatory review before reaching Chinese customers.

Data presents an even more difficult structural obstacle. Tesla’s neural-network approach improves by processing large volumes of video and driving information, but data collected by vehicles in China is subject to localization, cybersecurity and cross-border-transfer controls. Intelligent-driving systems also require access to licensed mapping services, while foreign companies cannot independently conduct the relevant surveying and mapping activities. Tesla’s partnership with Baidu gives it access to lane-level navigation and a qualified domestic mapping provider, but it does not remove all restrictions on how road data can be owned, processed or transferred. The company must either develop a sufficiently capable local training and validation system or obtain permission to use Chinese data within a tightly controlled structure. Hiring additional test engineers shows that Tesla is building local validation capacity, but recruitment should not be interpreted as evidence that nationwide regulatory approval is imminent.

The commercial stakes are significant because FSD could help Tesla generate software revenue and differentiate vehicles that increasingly compete on digital functions rather than battery range alone. Tesla has historically charged Chinese customers 64,000 yuan for the FSD package, while several domestic brands have included comparable Level 2 navigation features in lower-priced vehicles or offered them without a large separate fee. Tesla’s China-made wholesale volume, including exports, increased 37.8% year on year to 93,579 vehicles in July 2026, marking a ninth consecutive month of growth. That recovery gives Tesla a larger installed base to which it could sell software, but it does not eliminate pressure from BYD, Xpeng, Li Auto, Huawei-backed systems and other local competitors that are adapting more quickly to Chinese roads, consumer preferences and regulations.

China’s new framework should therefore be understood as a market-wide safety threshold rather than a rule designed solely to exclude Tesla. The standard’s drafting involved domestic automakers, technology providers, research institutions and foreign manufacturers, indicating that regulators want a unified route toward commercial deployment. Tesla still brings a large fleet, a major Shanghai production base, extensive AI experience and an established mapping partnership. Its disadvantage is that its global data-driven model must be rebuilt around Chinese compliance requirements. A full Level 2 rollout could still create meaningful software revenue, but Level 3 or Level 4 approval will require far more than adapting the existing American version of FSD. Tesla must demonstrate localized safety performance, a compliant data architecture, transparent driver supervision and reliable operation under China’s complex traffic conditions.