China Official: Autonomous Driving Fails Traffic Rules

Chinese regulators have issued a stark warning regarding the current state of autonomous driving technology. Ma Mingyue, a researcher at the Road Traffic Safety Research Center of the Ministry of Public Security, stated that self-driving systems still exhibit major deficiencies in complying with traffic laws.

This admission highlights a critical gap between marketing claims and operational reality in the automotive AI sector. The comments were made during the 2026 Future Car Pioneers Conference, which opened on May 29.

The revelation suggests that despite billions in investment, fully autonomous vehicles are not yet ready for unrestricted public road deployment. This stance aligns with previous warnings from Chinese officials about the limitations of current 'smart driving' features.

Key Facts

• Regulatory Stance: The Ministry of Public Security asserts that no commercially available system has achieved true autonomous driving capabilities.
• Human Liability: Drivers remain the primary legal responsibility holders for any incidents involving assisted driving systems.
• Tesla FSD Issues: Reports indicate Tesla's Full Self-Driving (FSD) initially struggled to recognize non-motorized vehicle lanes in China.
• Safety Shortboard: Inability to strictly follow traffic rules is identified as a major barrier to rapid industry expansion.
• Timeline Context: These comments follow earlier statements by Director Wang Qiang in July 2025 regarding system limitations.
• Market Impact: Major automakers and tech firms must adjust expectations for Level 4 and Level 5 autonomy rollouts.

Regulatory Scrutiny on Autonomy Claims

Ma Mingyue’s presentation at the conference served as a direct challenge to the aggressive timelines proposed by many Western and Chinese tech companies. He emphasized that while sensor technology and computing power have advanced, the logical decision-making processes of AI drivers often fail to interpret complex traffic environments correctly. This is particularly evident in mixed-traffic scenarios common in Asian cities.

The core issue lies in the interpretation of traffic rules. Current AI models rely heavily on pattern recognition rather than a deep understanding of legal obligations. For instance, an algorithm might prioritize efficiency over strict adherence to lane markings if it perceives a clear path. However, this behavior violates traffic codes designed to ensure predictable movement for all road users.

This regulatory pushback is not isolated to China. Similar concerns are being raised by the National Highway Traffic Safety Administration (NHTSA) in the United States and the European Union’s AI Act frameworks. Regulators globally are moving towards stricter certification standards for autonomous systems before they can be marketed as 'self-driving'.

The Tesla FSD Case Study

A specific example cited by Ma involved Tesla’s Full Self-Driving (FSD) beta software upon its initial entry into the Chinese market. The system reportedly failed to distinguish between motorized lanes and non-motorized vehicle lanes, such as those designated for bicycles and e-scooters.

Consequently, the vehicle attempted to drive in areas reserved for slower, vulnerable road users. This error underscores a fundamental limitation in current computer vision models. They struggle with context-aware navigation when faced with infrastructure that differs significantly from their training data sets, which are often dominated by North American or European road layouts.

Liability and Human Responsibility

The distinction between 'assisted driving' and 'autonomous driving' is legally crucial. Wang Qiang, Director of the Traffic Management Bureau of the Ministry of Public Security, clarified this position in July 2025. He stated that all currently sold vehicles with 'smart driving' systems require human oversight.

This means that in the event of an accident, the human driver is held liable, not the software manufacturer. This legal framework places a heavy burden on consumers who may trust the technology too implicitly. It also protects automakers from immediate product liability lawsuits related to software failures.

However, this dynamic is shifting as lawsuits mount globally. Plaintiffs in the US and Europe are increasingly arguing that marketing materials create a false sense of security. If companies name their features 'Full Self-Driving', consumers reasonably expect the car to handle all driving tasks without intervention.

• Current Legal Status: Human driver is the final responsible party.
• System Limitation: No commercial system meets SAE Level 5 autonomy standards.
• Marketing Risk: Aggressive naming conventions may lead to consumer misuse.
• Insurance Implications: Premiums may rise for users of advanced driver-assistance systems (ADAS).

Industry Context and Global Comparison

The situation in China mirrors global trends where technological hype outpaces engineering readiness. Companies like Waymo and Cruise in the US have faced similar scrutiny regarding their ability to navigate unexpected obstacles. While Waymo operates under strict geofenced conditions, broader deployment remains elusive.

Unlike previous generations of ADAS, which handled simple tasks like lane keeping, modern AI systems attempt complex urban navigation. This leap in complexity introduces unpredictable edge cases. A system trained on millions of miles of data may still fail when encountering a rare combination of weather, lighting, and unusual pedestrian behavior.

Western competitors like Mercedes-Benz have received approval for Level 3 autonomy in specific regions, but even these systems require the driver to take control when prompted. The Chinese regulator’s stance reinforces that true Level 4 or Level 5 autonomy, where no human intervention is ever needed, is still years away from mass-market viability.

What This Means for Stakeholders

For developers, the message is clear: robustness must precede scalability. Investing in better simulation environments and diverse training data is essential. Models must be tested against local traffic laws specifically, rather than relying on generic rule sets.

Businesses investing in autonomous fleets need to recalibrate their return on investment projections. The timeline for removing safety drivers is longer than anticipated. Operational costs will remain higher due to the need for remote monitoring and manual intervention teams.

Users should exercise extreme caution. Trusting an AI system completely can lead to catastrophic outcomes. Drivers must keep their hands on the wheel and eyes on the road, regardless of what the dashboard displays suggest. Education campaigns are needed to clarify the actual capabilities of these systems.

Looking Ahead

The path forward involves tighter collaboration between tech firms and regulatory bodies. Standardized testing protocols for traffic rule compliance will likely become mandatory. We can expect more frequent software updates focused on legal adherence rather than just feature additions.

In the next 12 to 24 months, we may see a shift in marketing language. Terms like 'Autopilot' or 'Full Self-Driving' could face regulatory bans if they do not accurately reflect the level of automation. This will help manage consumer expectations and reduce liability risks for manufacturers.

Ultimately, safety must trump speed. The industry’s credibility depends on demonstrating that AI can drive safer and more lawfully than humans. Until then, the human driver remains the ultimate safeguard on our roads.

Gogo's Take

• 🔥 Why This Matters: This confirms that the 'robotaxi' dream is delayed. Investors and consumers must accept that human oversight is not a bug, but a necessary feature for the foreseeable future. It validates the skepticism around rapid AI adoption in physical spaces.
• ⚠️ Limitations & Risks: The risk of 'automation bias' is high. Users may disengage from driving tasks because the system works 99% of the time, only to fail catastrophically in the remaining 1%. Legal ambiguity leaves victims of accidents in a difficult position regarding compensation.
• 💡 Actionable Advice: Do not trust marketing names. Treat any 'smart driving' feature as a sophisticated cruise control. Always maintain situational awareness. Compare systems based on safety records, not just feature lists. Wait for Level 4 certification before considering hands-off usage in complex urban environments.