Who is at fault if a driverless car crashes?

The general framework

Liability in an autonomous vehicle crash depends on three things:

1. What mode the vehicle was in. Fully autonomous, supervised (driver-monitored), or human-driven.
2. Who the operator is. The commercial entity running the AV service.
3. What state law says. Liability frameworks vary by jurisdiction.

For commercial robotaxi services like Waymo operating in fully autonomous mode, the operator is generally the responsible party. Not an individual driver, because there is no individual driver.

How it works in practice

For a crash involving a Waymo or Tesla Robotaxi in autonomous mode:

• The operator's commercial insurance carrier handles first-party claims (their own vehicle, riders) and third-party claims (other drivers, pedestrians, property).
• The operator is generally the named defendant in any civil litigation arising from the crash.
• Riders are typically covered as passengers under the operator's policy and any state-mandated coverage frameworks.

State-by-state variation

The legal frameworks differ across the states where commercial robotaxi services operate:

• California has explicit AV permit and operator-liability frameworks managed by the DMV and CPUC. See the permit glossary for the regulatory context.
• Arizona has historically taken a less prescriptive approach but operates under standard state vehicle-liability law applied to the operator.
• Texas uses a mix of state vehicle law and AV-specific statute.
• Other states with commercial AV service follow similar patterns; some have adopted model legislation, others apply existing fleet-vehicle frameworks.

NHTSA's role

NHTSA's Standing General Order on Crash Reporting requires AV operators to report qualifying crashes. This provides the federal-level investigation track separate from civil liability questions. NHTSA investigates patterns and can issue safety recalls; civil liability is handled separately through state courts and insurance carriers.

For more on what happens immediately after a crash, see what happens if a Waymo gets in an accident.

Tesla FSD (Supervised) is different

This is the place the public conversation gets muddled most often.

Tesla Full Self-Driving (Supervised) (the feature available on consumer Tesla vehicles in 2026) is an SAE Level 2 driver-assist system. The driver is legally required to monitor the road and be ready to take over. If a crash occurs:

• The licensed driver remains the legally responsible party, even if FSD was engaged.
• Tesla's product liability may be implicated if a defect contributed to the crash, but the driver is the primary respondent.
• This is fundamentally different from Waymo's situation, where there is no driver to hold responsible.

The Tesla Robotaxi pilot in Austin operates in unsupervised mode within its geofenced area. For trips in that mode, Tesla as the operator bears liability the way Waymo does. See is Tesla Robotaxi available for the operational distinction.

What if another driver caused the crash?

Most reported incidents involving Waymo vehicles fall into this category. A human-driven vehicle struck the Waymo. In these cases:

• The at-fault human driver (or their insurance) is the responsible party.
• Waymo's insurance handles the Waymo's repair and rider claims, then may subrogate against the at-fault driver's insurance.
• NHTSA's Standing General Order still requires reporting because the AV was involved, but the liability conversation is conventional.

This is why "Waymo was in a crash" and "Waymo caused a crash" are not interchangeable. See how many fatal crashes Waymo has had for the breakdown.

Bottom line

For an autonomous robotaxi in autonomous mode: the operator is liable, the operator's insurance pays, the operator is sued. For a Tesla on Supervised FSD: the driver remains the legally responsible party. NHTSA investigates patterns regardless of who's named in civil cases. State law governs the specifics.

This is the strict-bar Deploy answer; the public conversation often conflates the driver-assist and full-autonomy frameworks, which leads to confusion about who's responsible when something goes wrong. For methodology canonical references applicable to driverless crash liability framing: the 4-way autonomy-boundary taxonomy (driver-assist vs full-autonomy autonomy-boundary distinction) + verified-vs-claimed at within-entity granularity.