China’s Ministry of Industry and Information Technology (MIIT) has released a draft mandatory national standard for L3/L4 autonomous driving systems, with an implementation date of July 1, 2027. The proposed rules, titled “Intelligent Connected Vehicle Autonomous Driving System Safety Requirements,” mark a major shift from optional technical guidance to enforceable market-access standards for advanced driver-assistance and autonomy functions.
The most consequential element is a clear requirement for responsibility transfer. Under the draft, an L3 or L4 autonomous driving system must provide a takeover window of no less than 10 seconds after issuing a takeover request. If the driver does not respond in time, the system must execute a Minimal Risk Maneuver (MRM), guiding the vehicle into a safe stopping state. By putting a measurable time threshold and a defined safety outcome into the rules, MIIT’s draft aims to close the ambiguity that has accompanied earlier commercialization of higher-level driving assistance.
The standard also pushes manufacturers to move beyond launch-event demonstrations and toward certified safety products. In prior years, companies competed by showcasing capabilities such as urban Navigate on Autopilot and emphasizing high computing performance, while accountability in real-world scenarios could remain unclear. With the 10-second takeover deadline and the requirement for an MRM, the compliance burden becomes more explicit: L3 systems will need to be validated, documented, and certified under the new safety framework before they can be sold as L3 functions.
Safety-driven redundancy is expected to raise vehicle costs. Industry participants anticipate that meeting the standard will require additional hardware and safety layers, including redundant sensing coverage, dual or redundant computing approaches, and fail-safe control and braking systems. The draft’s demanding operational expectations are projected to increase per-vehicle L3 hardware costs by roughly 10% to 20%, narrowing the gap between the economics of L2+ and higher-level systems.
The new regulatory direction is likely to reshuffle competition among automakers and technology suppliers. Huawei is widely seen as particularly well-positioned to comply, supported by its full-stack approach and its role in shaping the regulatory direction. The company’s vertically integrated architecture—spanning its MDC computing platform, HarmonyOS software environment, and its perception and algorithm capabilities—is described as aligning well with safety certification logic.
In addition, Huawei’s emphasis on carrier-grade reliability and high-reliability system design may help in meeting functional safety and SOTIF-oriented certification requirements.
BYD, while strong in manufacturing scale and cost control, is framed as facing a different challenge in the L3 era: software systems engineering at the level required for mass certification. The key issue is less about building affordable hardware and more about establishing a full lifecycle safety loop—from research and development through testing, validation, and industrialization—so that safety development processes can withstand regulatory scrutiny and be repeatable at volume.
Newer EV makers such as NIO and XPeng face another set of constraints. With mandatory safety requirements, software update strategies can no longer rely solely on rapid iteration. Critical system changes may require re-certification or additional validation, making it harder to sustain “move fast” OTA cycles without triggering added compliance work. XPeng’s leadership has also signaled skepticism about L3 as a destination, emphasizing a path that jumps beyond L3 toward later autonomy levels.
International manufacturers are confronting their own tradeoffs. Tesla’s Full Self-Driving is still positioned in many markets as an L2 system, leaving a regulatory mismatch with L3 requirements that presume the system—rather than the driver alone—accepts defined responsibility during the autonomy operation window. In Germany, the commercial rationale for L3 has already weakened. Mercedes-Benz suspended rollout of its L3 Drive Pilot in early 2026, citing strict usage conditions and high costs relative to practical value. BMW likewise confirmed in March 2026 that its updated 7 Series would drop the L3 option due to low take rates, redirecting attention toward more broadly applicable L2+ offerings.
The draft standard also intensifies the “sensor architecture” debate. While MIIT’s requirements do not explicitly mandate any single sensor type such as LiDAR, the strict expectation that systems must continue to perform safely through edge conditions—bright light, heavy rain, dense fog—and even single-point sensor failures raises the bar for purely vision-based designs. At L3, the system must handle extreme scenarios within the autonomy window, and only a properly validated redundancy strategy can satisfy the MRM obligation if the driver cannot immediately intervene.
That pressure is already reflected in growing interest in high-performance computing and sensor fusion. Meeting the draft L3/L4 safety expectations is prompting demand for high-compute systems, with requirements often framed above 250 TOPS and some designs targeting 500 to 1,000 TOPS for advanced operation. As a result, NVIDIA (including platforms referenced as Orin/Thor), Qualcomm (Snapdragon Ride), and domestic players such as Horizon Robotics and Black Sesame Technologies are positioned as key suppliers for the compute stack.
Even with a formal regulatory path, L3 remains commercially constrained. L2+ systems continue to spread quickly due to lower cost and broader applicability, while L3 adoption is expected to remain limited to higher-priced vehicles—where customers may tolerate the incremental cost and where manufacturers can justify the investment in certified safety hardware. The draft therefore sets the stage for a compliance-driven transition, where engineers and suppliers must prove not only capability, but safety responsibility under the newly defined transfer-and-stop rules.
With MIIT’s mandatory standard scheduled for July 1, 2027, automakers and technology providers are expected to shift their focus from feature-first competition to certified systems engineering—prioritizing validation depth, safety accountability, and repeatable compliance processes across platforms and production lines.
