Mandatory National Standard GB 47955-2026 Deep Dive: Unified Legal Standards for L2 Autonomous Driving Data Forensics

The penetration rate of Level 2+ advanced driver-assistance systems (ADAS) in new vehicles in China now exceeds 70%. As NOA features rapidly become mainstream, a growing number of traffic accidents, dangerous driving incidents, and disputes between drivers and vehicles are occurring.Autonomous Driving System Operation, Driver Hands-Off / Distraction, System Intervention, and Pre-Collision DataIt is the core physical evidence for determining human-machine responsibility. For years, public security investigations and judicial appraisal industries have faced four major challenges in evidence collection:

• Inconsistent data storage formats and field definitions across automakers result in incomplete data extraction and an inability to perform comparisons.

• Lack of support from mandatory national standards may lead to challenges regarding the legality of evidence collection during trial cross-examination of appraisal reports.

• No unified data reading protocol; high difficulty in adapting multi-brand vehicle forensic tools.

• Critical events such as collisions, driver distraction, and system failures lack mandatory data retention locks, making raw data susceptible to overwriting.

1. On 2026-6-27, the national mandatory standard GB 47955—2026 "Safety Requirements for Intelligent Connected Vehicle Combined Driving Assistance Systems" was officially released. It will take effect on 2027-1-1. As China's first mandatory national standard for L2-level intelligent driving,Appendix F, "Data Recording Requirements," establishes a unified national mandatory baseline for in-vehicle electronic forensics.From data storage, record fields, read protocols to anti-overwrite mechanisms, it provides statutory and authoritative support for forensic scene investigations by public security organs and non-destructive examinations in judicial appraisals. This article fully dissects the core clauses of evidence collection from the perspectives of case handling, evidence gathering, and appraisal practices, based on the original standard text.

2. Standard draft for public comment. Please refer to the official final release for the final version:

Mandatory National Standard: "Safety Requirements for Combined Driving Assistance Systems for Intelligent Connected Vehicles"

I. Standard Basic Framework

1. Comprehensive coverage of vehicles and systems

Applicable to national standardsAll M-class passenger vehicles and N-class goods vehiclesCovers three categories of combined driver assistance systems: ① Basic single-lane (Lane Keep Assist + ACC); ② Basic multi-lane (driver-initiated lane changes); ③ Pilot-assisted combined driving assistance (NOA highway pilot).

2. Phased Implementation Rules

• New vehicle models declared starting from 2027/1/1: Must meet all requirements except for data recording.

• New vehicles with 13 months of implementation: Fully comply with all mandatory data recording requirements in Appendix F;

• Existing vehicle models covered by the announcement: Mandatory compliance data recording system after 25 months of implementation.

Evidence Notice: For new vehicle models launched after 2027 years, automakers are not permitted to simplify intelligent driving data storage. Public security authorities and appraisal institutions can directly require automakers to provide complete in-vehicle raw data in accordance with national standards. For existing older vehicles, data integrity should be verified according to standard requirements.

3. Core Forensic Evidence: Chapter F (Normative Mandatory Appendix)

Appendix FData Recording System for Combined Driving Assistance (DSSCDA)It is the core of full-text evidence collection, mandating that all Level 2 vehicles be equipped with standardized onboard black boxes. These are categorized into two types: Type I and Type II recording devices, with clear specifications.Critical Data, Storage Capacity, Write Protection, Standardized Read ProtocolFour hard rules.

II. Appendix F: Core Forensic Provisions

(I) Key Forensic Points for Distinguishing Two Types of DSSCDA Systems

1. Type I System (Mandatory for M1/N1 Passenger Vehicles)

   Lock records upon collision or high deceleration risk events.5s before and after event 1Complete time-series data: Timestamped events (activation, exit, hand-off alert, risk deceleration) are permanently retained. The system stores at least 5 collision events and 2500 event records, with locked data protected from overwriting.

   ✅ Evidence advantage: Critical pre-crash operation data is preserved after an accident and won't be overwritten by subsequent driving, so you don't need to worry about data loss.

2. Type II System (Commercial Vehicles + Optional Passenger Cars)

   Continuously records vehicle speed, steering, detected objects, and driver status for 8 hours without interruption, preserving complete system dynamic data. Ideal for retrieving full process logs in long-duration assisted driving dispute cases.

(II) Mandatory A-Level Core Forensic Data Required by National Standards (Core Basis for Appraisal Reports)

Data classified by national standards into Category A (mandatory recording for 100% of automakers) and Category B (recording only when equipped with the corresponding sensor).Class A data is the core evidence for finalizing a case.When the public security agency or appraisal institution collects evidence, they can verify completeness item by item:

1. Vehicle Basic Identity Evidence (Essential for Traceability)

VIN, DSSCDA Hardware Serial Number, Autonomous Driving System Software Version, UTC Timestamp (to the second), GPS Coordinates, Total Mileage

Forensic Value: Prevents data tampering by binding each log entry to a unique vehicle identity, enabling verification of original data source integrity during audits.

2. Vehicle Dynamic Driving Evidence (Core for Accident Reconstruction)

Real-time vehicle speed, lateral/longitudinal acceleration, yaw rate, and steering wheel angle. High-frequency recording at a minimum of 10 Hz to fully reconstruct braking, lane changes, and collision events.

3. Driver Behavior Evidence (Key to Distinguishing Human vs. Machine Responsibility)

• Hand-off and gaze-off states (complete HOR/EOR alarm logs);

• Accelerator/brake pedal position, driver steering input torque, seatbelt status;

Incident Investigation: In highway NOA accidents, directly retrieve data on hands-off duration and eye-closure distraction to determine whether the driver fulfilled monitoring obligations, distinguishing between system defects and human error.

4. Full logs of the intelligent driving system

System Activation / Deactivation, RM Risk Mitigation Parking Trigger, DCA Emergency Warning, AEB Emergency Braking Request, Various Lane Change Maneuvers (Driver-initiated / System-initiated).

5. Environmental Perception Evidence (Core of Forensic System Failure)

Target IDs, relative positions/velocities of vehicles ahead, pedestrians, two-wheelers, and obstacles; forward multi-view image/video (minimum 90 megapixels with millisecond timestamps);

Value Assessment: Reproduce road obstacles and jaywalking scenarios to evaluate whether the system exhibits perception failures or lacks warnings.

(3) Mandatory Tamper-Proof and Overwrite-Prevention Mechanism (Resolves Evidence Admissibility Concerns)

1. Event Locking Mechanism: Collision-related data is locked. Standard driving logs cannot overwrite locked accident records, preventing automakers from erasing accident history at the hardware level.

2. Synchronize to Unified UTC Time:All video, dynamics, and perception data across the vehicle share a single time source, enabling multi-dimensional cross-validation to eliminate timeline inconsistencies.

3. Non-volatile local storage:Non-volatile data retention ensures event logs are preserved even after vehicle submersion, power loss, or shutdown. Data can be recovered from the storage chip even if the vehicle is damaged in an accident.

(IV) Standardized Data Reading Protocol (Breaking Automakers' Access Barriers)

National standards mandate that automakers open three standardized vehicle data access channels: 11/29-bit CAN bus protocols, Ethernet read interfaces, unified data upload and CRC verification processes, and standard diagnostic read services.

✅ Solves Industry Pain Points: Previously, diverse proprietary encryption protocols across vehicle brands required appraisal agencies to purchase brand-specific forensic tools. With the national standard now in effect, a universal in-vehicle forensic device can legally extract DSSCDA raw data, significantly reducing inspection costs for multiple models. Police on-site investigations can now quickly read data without requiring technical support from automakers.

III. Practical Value for Police Operations

1. New Standards for On-Site Seizure of Vehicles Involved in Cases

After the accident vehicle is impounded, immediately seal the DSSCDA storage unit. Prohibit repeated power cycling and short-distance driving to prevent ordinary logs from overwriting locked collision data. The inspection record must specify the vehicle's power-on status and sealing time, and a video recording of the entire sealing process must be captured to establish the legality of the electronic evidence.

2. Retrieving automaker data is supported by mandatory regulations.

When law enforcement agencies request local DSSCDA data or cloud backups from vehicle manufacturers using legally issued documents, manufacturers cannot refuse on grounds of encryption or privacy. Since this is a mandatory national standard, it serves as valid written authorization for data requests; non-compliance may be referred to the market regulation authorities for action. Alternatively, third-party automotive forensics tools such as Zhongzhi Dingzheng J精灵 100 can be purchased to extract DSSCDA data.

3. Standardized Evidence Collection Checklist for Accident Liability Determination

Mandatory on-site data dimensions (suitable as an inspection checklist): ① Collision event: lock 1 seconds of pre-event sequence data; ② Complete records of driver hands-off and attention alerts; ③ System logs for lane changes, deceleration, and warnings; ④ Forward perception images and relative motion data with obstacles; ⑤ Vehicle dynamics curves including speed, braking, and steering.

4. Enhanced evidence collection for dangerous driving cases

For dangerous driving cases involving NOA misuse on highways, hands-off-wheel incidents, or phone usage while driving, the system can extract continuous DSSCDA records to capture the full duration of driver inattention and document multiple system alert warnings, forming a complete chain of evidence for violations.

IV. Core Benefits for Automotive Data Forensic Institutions

1. New statutory authoritative basis added to appraisal documents

When issuing a forensic identification opinion on electronic evidence for vehicles related to advanced driver-assistance systems (ADAS), direct citation is permissible.GB 47955—2026 Appendix FAs a national standard for digital forensics and data integrity verification, it significantly reduces challenges to evidence during court cross-examination, addressing the previous lack of a unified industry standard.

2. Standardize inspection processes and unify industry operating standards.

Accredited laboratories may develop standardized work instructions based on Appendix F:

• Hash verification of raw storage media images;

• Verify missing fields against the Class A mandatory data checklist item by item;

• Losslessly extract using GB-standard CAN/Ethernet protocol;

• Distinguish between Type I and Type II systems; separately extract lock events and logs for continuous 8 hours.

3. There are quantifiable criteria for determining human-machine liability.

When system failures or driver distraction disputes arise in accidents, experts can leverage GB-mandated data—such as hands-off duration, warning sequence timing, and detected target information—to quantitatively differentiate fault between parties. This objective, reproducible expert opinion enhances its admissibility in court.

4. Missing vehicle data may result in a defect determination

If the submitted vehicle's DSSCDA lacks mandatory A-level fields per national standards, the appraisal report can explicitly state non-compliance with GB 47955 mandatory clauses. This evidence of product non-conformity directly impacts accident liability determination.

V. Standardized Collaborative Case Handling Process for Public Security and Judicial Appraisal

• Front-end Public Security Investigation:Power off and seal the accident vehicle; record the VIN and sealing status; synchronize fixed in-vehicle video and dashboard footage; issue an electronic data extraction notice.

• Mid-tier appraisal and inspection:Transfer to a lab with in-vehicle electronic evidence certification. Perform lossless DSSCDA imaging per GB Appendix F, extract and hash the full dataset for integrity verification, and validate completeness of mandatory record fields.

• Backend Verification Issued:Reconstruct accident timelines per GB 47955 data record clauses, distinguish driver actions from system control behaviors, and issue a legally binding forensic report;

• Backend Judicial Acceptance:Public security, procuratorate, and judicial authorities may directly cite this mandatory national standard to certify the legality and validity of DSSCDA raw data as key electronic evidence for case adjudication.

GB 47955—2026: China's first mandatory national standard for L2 driver assistance systems, Appendix F: DSSCDA Data Recording SystemComprehensively address gaps in the intelligent connected vehicle electronic evidence system and end the industry-wide chaos of unregulated data recording and retrieval. For public security traffic management and criminal investigation departments, standardized onboard black boxes provide a regulatory basis for on-scene evidence collection. For automotive data judicial appraisal institutions, the national standard offers authoritative legal support for inspection procedures and expert opinions, significantly reducing litigation cross-examination risks.