Fully autonomous and conditionally automated vehicles are no longer hypothetical. Robo-taxi deployments in multiple major cities, Level 3 production vehicles from several manufacturers, and an expanding autonomous commercial fleet have moved AV crash litigation from law review articles to actual dockets. The plaintiff lawyer's challenge is that the liability framework, the insurance architecture, and the discovery toolkit for these crashes were all designed for human-driver incidents and fit AV crashes poorly. Understanding where the gaps are is now a practical competency for auto plaintiff counsel.
Automation Levels and Liability Allocation
The SAE International automation level taxonomy, codified in SAE J3016, provides the baseline for understanding who is responsible for vehicle control at the time of a crash. Levels 0 through 2 involve driver-assisted features where the human driver remains responsible for monitoring and overall control. Level 3 introduces conditional automation where the vehicle handles all driving tasks under specific conditions, but the driver is expected to be available to take over on request. Levels 4 and 5 involve high and full automation where the vehicle operates without human intervention in defined operational design domains (ODD).
The liability allocation question tracks the automation level. In a Level 2 crash, the human driver's conduct remains the primary liability focus; the vehicle's automated features are a parallel product defect theory. In a Level 3 crash, the liability question is more complex: was the automated system operating within its ODD when the crash occurred, and did it provide adequate transition demand notice to the driver? If the system failed within its ODD, the manufacturer bears primary liability. If the driver failed to respond to a proper transition demand, comparative fault analysis applies. In a Level 4 or 5 robo-taxi or commercial AV crash where there is no human driver at all, the manufacturer bears the full duty of care for the vehicle's operation.
OTA Software Update Liability
Over-the-air software updates have become standard across major vehicle automation platforms. Manufacturers push updates to vehicle control systems without requiring owner action, and those updates can materially change how the automated system behaves. When a crash follows an OTA update to the driving automation software, the causal relationship between the update and the crash is a central liability question.
Discovery in OTA update cases should seek: the complete version history of the vehicle's automation software, including the date and content of every update applied to the vehicle; the internal testing and validation records for the specific update version that was installed at the time of the crash; any known issue reports or rollback actions taken for that update version after the crash; and the manufacturer's adverse event reporting records with NHTSA for crashes occurring within a defined period following the update's deployment. NHTSA's AV-related standing general order requires manufacturers to report serious crashes involving automated driving systems; those reports are public records and often identify the software version and circumstances with enough specificity to confirm or rule out an update-related causal theory.
Data Preservation and the AV Black Box
Autonomous vehicles generate substantially more data than a standard EDR. A Level 3 or higher automated vehicle typically records: sensor data from cameras, LIDAR, and radar arrays; the autonomous system's real-time decision log showing what the system detected, what it decided, and why; system state logs showing whether the automation was engaged and within its ODD; and driver monitoring data where the platform includes occupant attention monitoring. This data is the AV equivalent of both an EDR and a flight data recorder, and it is far more detailed than anything available in a standard crash investigation.
Send preservation demands to the manufacturer immediately after an AV crash. Cloud-synchronized data platforms are common in AV systems: the vehicle uploads trip and event data to the manufacturer's servers after each trip, which means a preservation demand to the manufacturer addresses data that may no longer be on the vehicle. State the specific data categories by name, reference the crash date and vehicle VIN, and send the demand by means that create a timestamped delivery record. Manufacturers have litigation holds that are triggered by notice; the hold protocol does not activate automatically from the crash event alone.
The Insurance Coverage Gap
AV crashes expose a coverage gap that standard auto policies were not written to address. Many standard auto policies provide that coverage applies when the insured is operating the vehicle. In a Level 4 or 5 crash where the vehicle was operating autonomously with no human input, defense counsel for the insurer may argue the insured was not operating the vehicle and that the vehicle manufacturer's product liability coverage, not the auto policy, is the applicable source of coverage.
Several states have enacted AV-specific legislation that addresses insurance requirements: California, Florida, Texas, and Michigan all have statutory frameworks that address liability and insurance requirements for autonomous vehicle deployment. The California AV testing permit regulations administered by DMV require proof of insurance or self-insurance as a condition of deployment. Where a robo-taxi company is the operator, its commercial general liability and auto liability policy structure is the primary coverage source; the vehicle manufacturer's product liability policy is a secondary source for defect claims.
For coverage of ADAS failure as a product defect theory in passenger cars with driver-assisted features, see our auto accidents practice area. Product defect and federal preemption analysis for vehicle software and control systems is addressed in our product liability section.