October 9, 2026 · 10 min read

ISO 26262-3 §6 / -5 §7 / -9 §7ISO 16750-4 / ISO 20653 / UN R79NHTSA 26V621 / PE26-002

When the Connector Was the Single Point of Failure: Ram ProMaster 26V621 and the EPS Harness That Corroded Through an ISO 26262 Lens

Electric power steering is a safety-critical function that most drivers only notice when it is gone. On a loaded cargo van, that means the wheel suddenly weighs what it did before assist was invented. The cause, according to the filing, is not a clever failure mode. It is water in a connector.

The interesting part is not that a harness corroded. Harnesses corrode. The interesting part is the ratio between what the regulator heard and what the manufacturer already held in its own warranty database, and what that ratio says about where the safety case should have been looking.

1. The public record

The recall. FCA US (Chrysler) filed NHTSA campaign 26V621 (manufacturer number 92D) on September 29, 2026, after determining a safety defect on September 22, 2026. It covers 265,512 Ram ProMaster vans, model years 2022 to 2026: 44,391 1500s, 147,204 2500s, 72,043 3500s and 1,874 model-year 2024 to 2026 3500 EVs. The component is the body and jumper wiring harness for the electric power steering. The defect statement: "Corroded wiring harnesses can cause intermittent or permanent loss of power steering assist." FCA estimates 7.7 percent of the population has the defect. The filing says the EPS warning lamp may illuminate when assist is lost, but no warning before the failure is identified (NHTSA Part 573 report, 26V621).

The field data. FCA reports one potentially related accident with no injuries. It also reports 2,799 warranty claims, 437 customer assistance records and 191 field reports received from April 2022 to June 2026, plus 974 further reports as of September 2, 2026 (same filing).

The earlier signal. NHTSA opened Preliminary Evaluation PE26-002 on March 2, 2026, covering 136,666 model-year 2022 and 2023 ProMasters, on 19 Vehicle Owner Questionnaires. The stated problem was "Loss of power steering assist due to water intrusion into power steering module connector." The resume recorded no crashes, injuries or fatalities (NHTSA ODI resume, PE26-002). Trade coverage noted one owner video showing heavy corrosion inside a connector, with the wheel fighting the driver as assist cut in and out (The Autopian).

The timeline FCA filed.

| Date | Event | |---|---| | Apr 2022 | Earliest warranty and field reports in the filed data set | | Sep 22, 2025 | NHTSA pre-investigation inquiry opens on nine complaints | | Jan 8, 2026 | FCA internal investigation opens | | Mar 2, 2026 | NHTSA opens PE26-002 | | Jun 9, 2026 | NHTSA information request (FCA responds Jul 24) | | Sep 22, 2026 | FCA determines a safety defect exists | | Sep 29, 2026 | Part 573 report submitted | | Nov 17, 2026 | Interim owner letters begin; remedy not yet available |

The harnesses were supplied by Lear (Valls, Spain) and Flex (Ciudad Juarez, Mexico). The remedy is inspection and, if necessary, replacement of the harness and the EPS gear, with improved sealing against water. Repair launch is expected in the fourth quarter (Mopar Insiders; Forbes, October 7, 2026; Fox Business, October 7, 2026).

The arithmetic. My numbers, from the public ones:

2,799 + 437 + 191 = 3,427 manufacturer-side records, against 19 owner complaints in NHTSA's file. That is roughly 180 to 1. The regulator saw the tip; the manufacturer's databases held the iceberg.
7.7 percent of 265,512 is about 20,400 vehicles estimated defective. Against 2,799 warranty claims, roughly one in seven of the vehicles FCA expects to be defective had shown up by mid-2026. The rest are latent: a corrosion population that is still maturing.
From the first filed field data in April 2022 to the defect determination is about 4.4 years. From FCA's own internal investigation to the determination is 257 days; from the PE opening, 204 days.
The recall population (265,512) is 94 percent larger than the PE population (136,666). The failure did not stay in the years it was first noticed.

I am not suggesting anyone sat on information. I am saying the process ran at the speed of an investigation, and the corrosion ran at the speed of weather.

2. The standards lens

ISO 26262-3 §6 (HARA). Loss of steering assist is a textbook item-level hazard. Severity is arguable: with assist lost, a vehicle remains steerable with higher effort, so many programs rate S2 to S3 depending on speed and vehicle mass; a fully loaded van at highway speed does not belong at the bottom of that range. Controllability is the contested field. For a sudden, intermittent loss, controllability is worse than for a clean, latched, warned loss, because the driver cannot form an expectation. Exposure is effectively every driving state. The result is a safety goal along the lines of "avoid unintended loss of steering assist without driver warning," with an ASIL that the public record does not show. I will not guess it; I will note that the filing's central phrase is "no identified warning," which is a statement about the safety mechanism, not the hazard.

ISO 26262-5 §7 (hardware architectural metrics) and Annex D. A connector between the controller and the harness is a hardware part with a failure rate and failure modes: open, high resistance, intermittent contact, short to a neighbouring pin. Whether those modes were in the FMEDA, and with what rate, is not public. The point of the standard is that "connector" is not a free line item. Treating the harness as passive wiring with a nominal FIT is exactly how a corrosion mechanism escapes quantification.

ISO 26262-9 §7 (dependent failure analysis). Here is the core issue. Water ingress is not a random fault with independent occurrence per vehicle. It is a common-cause mechanism: the same seal design, the same underbody location, the same geography (road salt, washdown, humidity), and a monotonically increasing hazard rate with age. A model that treats it as independent per-hour random failure will underestimate the population risk, and will not predict that the failure shows up in clusters. Dependent failure analysis exists to ask which stressors couple many parts at once. Water is the first one on any underbody list.

ISO 16750-4 and ISO 20653 (environmental testing and ingress protection). These are where the environmental assumption is made testable: climatic loads including salt mist and humidity cycling, and an IP code for the connector assembly. The question is not whether the connector had an IP rating; it is whether the validation profile matched a van that is washed, parked outside, and driven in winter for ten years, with the connector location in the wheel-spray zone that the owner video suggests. A pass on the lab profile and a field failure at 7.7 percent is the signature of an environmental profile that did not represent the use.

ISO 26262-4 §6 and §8 (technical safety concept, safety validation). The safety mechanism question is the fault reaction. A well-designed EPS that detects a connector fault should (a) annunciate it, (b) reduce assist gracefully rather than abruptly where the hazard analysis calls for it, and (c) latch or debounce so the driver sees a stable state. The filing says there is no identified warning before the failure, and the lamp comes on when assist is lost. Intermittent contact is the hard case, because a monitor tuned to reject glitches will also reject the early symptoms.

ISO 26262-2 §7 and -7 (field monitoring, safety anomalies). The inquiry in September 2025 and the internal investigation in January 2026 are the moments a field-monitoring process turns warranty data into a safety review. The warranty claims had been accumulating since April 2022. The artifact that links claim volume to a safety trigger, with a threshold, is the one I would want to see.

Regulatory note. In UNECE markets, steering systems fall under UN R79, which sets requirements for warning and residual steering ability on failure of the assist. I am not aware of an equivalent US federal standard that mandates a specific EPS failure warning, so for a US-market van the requirement was largely self-imposed, and the safety argument lived in the manufacturer's own documents.

3. A worked snippet

An illustrative fault tree and FMEA built from the public record. These are mine, not FCA's.

TOP: Unwarned loss of steering assist at speed (Safety Goal SG-EPS-01)
 |
 +-- AND
      |
      +-- Assist torque lost or intermittent
      |    +-- OR
      |         +-- EPS module connector: high-resistance or open (corrosion)   <-- 26V621
      |         +-- Jumper harness conductor: open or intermittent (corrosion)  <-- 26V621
      |         +-- Motor or ECU internal fault
      |         +-- Supply undervoltage
      |
      +-- Driver not warned in time
           +-- OR
                +-- Fault debounce rejects intermittent contact
                +-- Warning raised only after assist is already lost
                +-- Lamp circuit shares the failed supply or connector

| ID | Failure mode | Effect | S | O | D | AP | Current control | Action | |---|---|---|---|---|---|---|---|---| | FM-01 | Connector contact resistance rises (corrosion), assist intermittent | Wheel effort changes abruptly | 9 | 6 | 7 | High | Debounced fault monitor | Add contact-resistance diagnostic; shorten debounce | | FM-02 | Connector open, assist permanently lost | Manual steering only | 9 | 5 | 6 | High | Lamp on loss | Pre-failure degradation warning | | FM-03 | Warning lamp not lit on intermittent loss | No driver awareness | 9 | 5 | 8 | High | None identified | Latch intermittent events, lamp on first occurrence | | FM-04 | Harness seal fails after wash and freeze cycles | Water reaches pins | 9 | 6 | 8 | High | Lab ingress test | Add field-profile cycle test; sealed connector |

S, O and D follow the AIAG-VDA 2019 scales; Action Priority is read from the AP table, not from a product of the three numbers. The ratings here are mine and exist to show the shape of the analysis.

4. Derived requirements (excerpt)

| ID | Requirement | Trace | |---|---|---| | SG-EPS-01 | The item shall avoid unwarned loss or intermittent loss of steering assist. | HARA row, ISO 26262-3 §6 | | TSR-EPS-014 | The EPS ECU shall detect connector contact resistance exceeding the calibrated limit and raise a driver warning within 500 ms of the first occurrence, with no debounce that suppresses a single event. | SG-EPS-01, FM-01, FM-03 | | TSR-EPS-015 | On any detected assist loss or degradation, the item shall latch the warning for the remainder of the ignition cycle and log a freeze-frame record. | SG-EPS-01, FM-03 | | HSR-EPS-031 | The EPS connector and harness interface shall remain functional after the vehicle-level salt mist, humidity and wash-freeze profile defined from fleet usage, validated per ISO 16750-4 and rated per ISO 20653. | FM-04, ISO 26262-9 §7 | | SM-FLD-007 | A field-monitoring rule shall open a safety review when EPS-related warranty claims exceed a defined rate per thousand vehicles in service for any build window or supplier lot, with the threshold stated in the safety plan. | ISO 26262-2 §7 |

Every row traces up to a safety goal and down to a verification method, which is the whole point of a requirement having an ID.

5. What the headline really tells us

The headline says water got into a wiring harness. The record says something more precise: a function whose loss is a steering event was being supported by a connector whose environmental failure mode was a common-cause, age-dependent mechanism, and the evidence that it was happening sat in the manufacturer's warranty data for years before it reached a safety decision. The missing artifact is not a better seal, although there will be one. It is an environmental dependent-failure analysis for the EPS interface, paired with a quantitative field-data trigger that converts claim volume into a safety review before the regulator opens a file.

Hardware that sits between a controller and a motor does not get to be an afterthought in a safety case. It is the single point of failure that the architecture quietly assumed away.

Sources

— Jherrod Thomas, The Lion of Functional Safety™