Nine days after three brand-new luxury MPVs fractured their brake pedal brackets on camera, the manufacturer's answer arrived in the form of a number twice as large as the one that broke the parts. Zunjie circulated a third-party test document on October 13 showing that 14 brake pedal assemblies withstood a stepped load of up to 3,000 newtons without cracking, roughly double the 1,500 newton benchmark in China's recommended pedal standard and comfortably above the 1,612 newtons recorded at the moment of fracture in Dongchedi's test. The report topped Weibo's search rankings, JAC Motor shares rallied as much as nine percent intraday, and the company's dealers began phoning owners one by one.
This is the phase of a quality dispute where narratives harden into camps, and the evidence on both sides is genuinely informative. The bench test is not a publicity stunt, and neither is the vehicle test we examined in our original analysis of the fracture. Understanding precisely what each measured, and what neither can measure, is the analytical work that matters for anyone assessing the risk to Zunjie, to its manufacturing partner JAC, and to the wider validation regime that Beijing rewrote two days after the original video went viral.
Inside the 3,000-Newton Test: What CAERI Actually Measured
The document, titled as an explanation of the V800 brake pedal inspection, states that testing took place on October 11, 2026 at a nationally recognised third-party institution. Multiple outlets, including Securities Times and Yicai, identified that institution as the testing centre of China Automotive Engineering Research Institute, or CAERI, a listed company trading in Shanghai under 601965. Report number 26-WT-DGJ-N10440 was confirmed by CAERI's own archive staff and by a JAC insider. The inspection followed the industry standard QC/T 788-2018 on pedal assembly performance and bench test methods.
| Dimension | Dongchedi vehicle test (Oct 8) | CAERI bench test (Oct 11) |
|---|---|---|
| Specimen | Three complete, unmodified new vehicles including one with roughly 3 km on the odometer | 14 brake pedal assemblies submitted by the manufacturer, four of them in the standard-load phase |
| Loading mode | Dynamic: consecutive full-force stops from 100 km/h, force applied through the pedal | Quasi-static: 20 mm per minute, load held five seconds then released, five cycles |
| Load level | Fracture at roughly 1,596 N and 1,612 N on the 2nd to 4th stop | 1,500 N standard, then stepped from 2,000 N to 3,000 N |
| Result | Bracket fractured in the plastic reinforcement around the pedal pivot, foot brake lost | No cracks or damage; mean permanent deformation 2.61 mm against a 5.0 mm limit, maximum 4.07 mm |
| Validity boundary | Real vehicle and real loading path, but a single self-designed protocol with no regulator or manufacturer present | Accredited laboratory and documented data, but CAERI says it vouches only for the submitted samples |
Two numbers deserve to be read carefully. The first is 2.61 mm, the mean permanent deformation across four samples that received the standard 1,500 newton load. That is a compliant result, but it is not a trivial one: permanent deformation means the pedal assembly took a set, and the standard's ceiling of five millimetres leaves roughly two millimetres of headroom on the average sample and less than one millimetre on the worst sample at 4.07 mm. A part that deforms this close to its allowance under the benchmark load is a part whose margin is thin rather than comfortable. The second number is 3,000 newtons, which does demonstrate that a properly manufactured assembly of this design does not have to fail at 1,612 newtons.
The Loading-Rate Problem: Why 3,000 N Static and 1,612 N Dynamic Can Both Be True
The disagreement is not about honesty, it is about mechanics. A bench test that pushes a pedal at 20 millimetres per minute and holds the load for five seconds characterises quasi-static strength. Real emergency braking delivers a dynamic transient: the load rises in milliseconds, the driver's foot and leg act as a spring, and the pedal assembly experiences bending and shear that a linear actuator does not reproduce. Dongchedi's own commentary made this distinction explicit, and independent experts including Li Yanwei of the China Automobile Dealers Association's expert committee noted that the published chart is a report summary which cannot be used to verify the complete test process.
The material specification makes the distinction sharper rather than weaker. The bracket uses polypropylene with roughly 40 percent glass fibre reinforcement, a composite chosen for mass, cost and corrosion performance. Glass-filled polypropylene is viscoelastic: its stiffness and failure behaviour depend on temperature, humidity and, critically, loading rate. Its failure mode is brittle rather than ductile, it is sensitive to weld lines and rib geometry, and its fatigue life under cyclic loading is not inferred from a static test at all. A quasi-static pass at 3,000 newtons is consistent with a dynamic fracture at 1,612 newtons under repeated cycles, which is exactly why the two datasets can coexist without either being fraudulent.
- Load path versus material strength: the fractures were in the plastic reinforcement wrapping the pedal pivot, which suggests a design geometry question rather than a material batch question.
- Cycle count versus peak load: the failure appeared on the second to fourth consecutive stop, the signature of fatigue accumulation rather than single-event overload.
- Sample provenance: CAERI tested assemblies the manufacturer provided; nothing published establishes that those units are drawn from the same lots as delivered vehicles, or that production process control is documented to that standard.
- Validation completeness: a defensible programme would run dynamic cyclic pedal loading at vehicle level, across temperature extremes, with instrumented fracture detection, and would publish the sampling plan.
Who Wins and Who Loses in a Dispute That Refuses to Settle
The proximate winner is Zunjie's short-term narrative control. Weibo's top search term, a dealer script with an accredited laboratory attached, and a nine percent intraday share rebound at JAC on October 13 are meaningful for a company whose customers are paying between 766,000 and 1,016,000 yuan for the vehicle. Delivery notices that had been postponed pending the pedal upgrade can resume, and sales staff who spent a week telling buyers that they had never seen a fracture now have a document to hold up.
But the maintenance costs are real. Expanding the free upgrade from the V800 to the S800 collectors edition, as at least one owner has documented, converts a single-model remedy into a component-family remedy, which raises cost and quietly concedes that the exposure is not confined to one nameplate. CAERI carries its own reputational exposure: a national institute whose name has been attached to a manufacturer-side document now needs to be understood as having said the report covers only submitted samples and that it cannot verify whether circulating copies were altered. Dongchedi, for its part, has shifted from publisher of a test to litigant in a public argument, having issued four consecutive social posts challenging the report, and its editorial position is now inseparable from the outcome. Buyers who paid deposits have already shown they read this as a trust question rather than an engineering one, with reports of customers refusing delivery and considering resale.
The Reality Check: What a Report Summary Cannot Prove
Three gaps separate the October 13 document from a settlement. The first is the difference between a summary and a report. What circulated is an explanation document built on a report summary, with data tables but no complete test record. Li Yanwei's point is precise: a summary evidences that the submitted parts behaved in a specified way under specified loading, nothing more. Independent access to the raw record, the sampling plan and the calibration chain is what would convert a claim into a finding.
The second gap is conformance between the tested samples and the delivered fleet. CAERI's own archive staff stated plainly that the institution is responsible only for the samples or vehicles the company provided and offers no guarantee for subsequently mass-produced products, while noting it cannot vouch for the integrity of the circulated document. For an OEM, that is the entire engineering question: not whether a hand-selected assembly can pass 3,000 newtons, but whether every unit built is equivalent to it. Nothing published to date addresses batch consistency, tooling, gate location, regrind content or the weld-line position that would drive a pivot-area fracture.
The third gap is the absence of an independent, regulator-visible replication of the dynamic condition. Both sides have now produced controlled tests that favour them: a single self-designed vehicle protocol that produced three failures, and a manufacturer-commissioned bench protocol that produced zero. Neither was conducted under a regulator's observation, and neither has been reproduced by a third party on sealed vehicles drawn from the delivered fleet. Under China's regulation on the recall of defective automobile products, that is the evidence path that determines whether an unreasonable safety risk exists, and it is the path that has not been walked.
A final observation on scale. Zunjie has delivered on the order of two thousand V800 units, and the argument for a recall does not require the defect to be universal. Three fractures across three vehicles from different production batches, at the same location and in the same failure mode, is a repeatability signal that regulators treat as material. A bench pass overrides that signal only if it explains the failure mode, and so far neither party has offered a mechanism: not the manufacturer, not the laboratory, and not the tester.
Regulatory and Export Consequences
The context has changed since October 8. On October 10, four ministries published draft rules requiring newly filed models to complete structural reliability verification that explicitly covers braking system components under cyclic loading, alongside at least 30,000 kilometres of reliability testing and a one-year environmental validation cycle covering two winters and one summer, with non-compliant filings barred from approval from January 2027 and existing approvals required to file materials by July 2027. Zunjie's dispute is now the reference case for that framework: a brake pedal bracket that passed a static standard and failed a cyclic dynamic one is precisely the scenario the draft rules were written to catch.
The standards themselves already contain the relevant hook. GB 21670-2025 requires that where failure of the brake pedal or its bracket would prevent a vehicle from achieving the required emergency braking performance, the part must be made of metal or an equivalent-performance material and must not visibly deform in normal operation. GB 7258-2026, effective from July 2027, caps driver pedal force at 500 newtons but does not test bracket capacity. QC/T 788-2018 remains a recommended rather than mandatory standard, which is why a compliant bench result and a fractured bracket are not legally contradictory. For export programmes, the implications run through European type approval and its durability expectations, where a documented dynamic fatigue result carries more weight than a static peak.
Precedent suggests how this ends when regulators engage. Ford China recalled 2,627 imported Mustang vehicles in 2020 because the pedal bracket material was changed from nylon to polypropylene without adequate design margin, producing fracture at the pedal pivot, and the remedy was a free replacement with an improved bracket assembly. The template is well established: identify the mechanism, define the affected population by batch, and replace the part under a filed recall rather than an opt-in service campaign.
Strategic Outlook: Three Scenarios for the Pedal Dispute
The dispute now turns on whether an independent dynamic replication happens before the November upgrade begins.
Bull Case
Zunjie files the full CAERI report, publishes the sampling and batch traceability data, names the upgraded material and opens the replacement to every affected VIN with a documented completion deadline. The dispute ends as a case study in transparent remediation, JAC recovers its valuation gap, and the November retrofit is completed with few unaddressed vehicles.
Base Case
The bench report plus the November upgrade under a service-campaign label satisfies most owners and the story fades, while the missing dynamic replication leaves a permanent question mark. Some owners never receive the upgrade because it remains opt-in, and the component family stays exposed to any future incident.
Bear Case
A fourth fracture in normal driving, or a regulator opening a defect investigation that demands sampling from the delivered fleet, forces a recall under conditions where the company has already spent its credibility on a static test. The reputational cost then falls on Huawei's alliance model as well as on the manufacturer, and JAC's next MPV programme carries a validation discount.
Five takeaways for executives and investors:
- Read the loading rate, not the headline load. A 3,000 newton quasi-static pass does not predict behaviour under dynamic transient braking, and the material specification makes that gap wider.
- Track the sampling plan, not the pass result. Whether the tested assemblies are equivalent to delivered units is the decisive engineering question that has not been answered.
- Treat the S800 expansion as a scope signal. Extending the remedy to the flagship sedan indicates a component-family exposure and increases the case for a formal recall filing.
- Watch for an independently witnessed dynamic test. A regulator-observed replication on sealed vehicles would settle the mechanism question in days; its absence is itself informative.
- Price the regulatory second-order effect. The October 10 draft rules make cyclic-load verification of braking components an approval requirement, so this case sets the enforcement benchmark for every Chinese manufacturer filing after January 2027.