China's MIIT Issues First Mandatory National Standard for L3/L4 Automated Driving Safety, Effective July 2027

Core Highlights

The Ministry of Industry and Information Technology (MIIT) has organized the development and formal approval of China's first mandatory national standard for automated driving safety, GB 44721-2026, 'Safety Requirements for Automated Driving Systems of Intelligent Connected Vehicles,' scheduled to take effect on July 1, 2027. This is the country's first compulsory national standard specifically targeting L3 conditionally automated driving and L4 highly automated driving systems. In practical terms, it establishes a unified, quantifiable, and enforceable safety-access baseline for advanced automated driving products in China. Until now, the domestic regulatory environment relied heavily on voluntary guidelines and fragmented local pilots, which left both automakers and consumers without a single authoritative yardstick for what 'safe enough' actually means. By elevating the document to mandatory status, the regulator signals that the era of loosely supervised high-level automation is closing.

The standard's arrival also matters because China is simultaneously the world's largest vehicle market and one of the most aggressive proving grounds for intelligent driving features. Dozens of brands now ship L2-plus systems, and several have begun limited L3 pilots under regional permissions. A national baseline prevents a race to the bottom in which companies cut safety corners to deliver flashier demos. Put differently, the standard converts a sprawling, uneven innovation landscape into a structured compliance regime that still leaves room for competition on engineering quality rather than on marketing claims.

Specific Capabilities and What Happened

The standard is an upgrade of the 2024 recommended (voluntary) national standard GB/T 44721-2024. The shift from 'recommended' to 'mandatory' marks a substantive change in the regulatory posture: previously, automakers could voluntarily reference the document, but going forward they must satisfy its hard requirements before a product can be sold. The standard covers safety requirements across the critical links of an automated driving system—perception, decision-making, execution, redundancy, human-machine interaction, and failure response—and defines the baseline metrics that products must meet during design, testing, and validation.

Beyond the headline change, the approval process itself reflects growing institutional maturity. MIIT coordinated input from vehicle manufacturers, tier-one suppliers, testing laboratories, and academic institutes, which means the eventual clauses are unlikely to be technically infeasible. The 2027 effective date also gives the industry a roughly two-year runway to redesign hardware architectures and revalidate software stacks. That lead time is deliberate: forcing immediate compliance would strand billions of yuan in already-tooled production lines, whereas a phased deadline lets incumbents adapt while still signaling seriousness.

Technical Details

A mandatory national standard carries legal force, distinguishing it from industry association standards or enterprise-defined specifications. GB 44721-2026 imposes systematic constraints on functional safety, safety of the intended functionality (SOTIF), cybersecurity, and data security for both L3 and L4. For example, when the system fails or operates beyond its operational design domain (ODD), the vehicle must be able to degrade safely, prompt the driver to take over, or enter a minimal risk maneuver (MRM) if no takeover occurs. This requires redundant braking, redundant steering, and redundant power supply capabilities.

Operationally, the document is expected to define how a system must behave across a matrix of edge cases: sensor occlusion in bad weather, conflicting perceptions, communication loss with infrastructure, and unexpected road users such as jaywalkers or animals. It also addresses the handover problem central to L3, where the driver is permitted to disengage from continuous monitoring yet must be able to resume control within a bounded time window. The standard therefore specifies driver-attention monitoring, takeover-request timing, and the fallback trajectory the car must follow if the driver fails to respond. These are precisely the failure modes that have made regulators worldwide cautious about conditional automation.

Comparison with Competing Frameworks

Globally, the UN WP.29 framework—through regulations R155, R156, and R157—has already established international baselines for automated driving and software updates, while the EU and individual US states have their own approval pathways. China's mandatory national standard echoes these efforts but is more centralized and uniform in its rollout: the access threshold is locked in at once through a national standard rather than being tested first by scattered local pilots. For domestic automakers, this reduces compliance uncertainty.

There are real differences in philosophy. The European approach leans on type approval and post-market oversight, while the United States has deferred heavily to state-level motor-vehicle codes, producing a patchwork. China's model—a single mandatory GB that applies nationwide—offers predictability but also concentrates risk: if the thresholds are set too high, smaller players are squeezed out; if too low, public confidence could erode after the first serious incident. The upgrade from the 2024 voluntary version suggests regulators chose an incremental path, learning from early industry feedback before imposing binding obligations.

Industry Impact and Applicable Scenarios

Simply put, this national standard is the examination syllabus for issuing 'driver's licenses' to advanced automated driving. For automakers and component suppliers, the development cycle before July 2027 will revolve around compliance: perception redundancy, fail-safe capability, and data closed-loop ability will become R&D priorities. For consumers, advanced driver-assistance and automated driving functions launched after the standard takes effect will be better protected in terms of safety. For the industry, a unified baseline helps curb excessive 'PPT autonomous driving' hype, allowing companies with genuine systems-engineering capability to stand out, and provides an institutional foundation for insurance, liability determination, and accident traceability.

The ripple effects extend well beyond carmakers. Insurers will gain a clearer basis for pricing policies on vehicles with higher automation, and courts will have a statutory reference when assigning fault between driver and system. Testing and certification bodies are likely to see a surge in demand for validation services, creating a new compliance economy. Long term, a credible safety baseline is a prerequisite for public trust, and public trust is what will ultimately determine how fast L3 and L4 features move from flagship trim levels into the mass market. In that sense, GB 44721-2026 is less a constraint than a foundation for the next decade of Chinese intelligent mobility.