If an Accident Happens When No One Is at the Wheel, Who Is at Fault?

Autonomous Driving and Artificial Intelligence
Author: Lütfi Mete Odabaşı
Creation: 12 October 2026

When a crash occurs while the steering is empty, responsibility now falls not on the driver's desk, but on that of the company that developed the system.

Share This Page

If a vehicle crashes with no one at the wheel, the one at fault is often not the driver but the company that developed the system: In Level 3 and above autonomous systems, legal liability shifts to the manufacturer, but in practice, who pays is divided among the driver, manufacturer, software developer, and insurer depending on the circumstances of the crash. In short, there is no one-word answer; the court files are speaking.

Highlights

$243 million: The jury found Tesla 33% liable in the fatal Autopilot crash in Miami.
Mercedes is the first Level 3 manufacturer Drive Pilot to assume legal liability if a crash occurs while the system is active.
More than 1,300 Tesla driver assistance system crashes were identified in US investigations.
Some US states require at least $100,000 in liability coverage for driverless vehicles.
Human error accounts for approximately 90-94% of crashes in the US.

Let's Clarify One Thing First: Is the Car Really "Autonomous"?

Tesla'nın FSD'si dahil yollardaki çoğu sistem Seviye 2'de: araba direksiyonu tutar ama kaza anında sorumluluk hep sürücünün kalır.

Not everyone means the same thing when they say "fully autonomous car," and the law is built precisely on this distinction. On SAE's scale of levels ranging from zero to five, most systems we see on the roads today, including Tesla's Full Self-Driving and GM's Super Cruise, are at Level 2. That means these are driver assistance systems: the car holds the steering wheel but keeps its eyes on you, and you are expected to take over at any moment. If a crash occurs, responsibility overwhelmingly falls on the driver.

Level 3, however, is a completely different game. At this stage, called "conditional automation," the vehicle takes over all driving tasks under certain conditions; the only thing required from the driver is to be ready when the system asks for control back. And this is exactly where the legal liability table flips: if a crash occurs while the system is driving, responsibility lies with the manufacturer.

But what about that "handover" moment in between? That is precisely where lawyers furrow their brows the most. Does responsibility return to the driver from the second the system says "take over," or does it stay with the manufacturer until the transition is complete? These blurry handoff scenarios created by Level 3 systems are the gray area that determines the fate of lawsuits.

Mercedes Drive Pilot: The First Brand to Say "If There's a Crash, I Pay"

So far, the only major player that has explicitly assumed liability is Mercedes. Drive Pilot While it is active, the driver is not legally responsible for the operation of the vehicle unless the system shuts itself down; if a crash occurs, the bill goes to Mercedes. If they crash after the system hands control back to you, that's another story, and responsibility returns to you. The company has also openly stated that it is the first automaker to receive international certification for a Level 3 system.

We are the first automaker to receive international certification for a Level 3 system.
Gregor Kugelmann, senior development manager for Drive Pilot

But wait, this is not an automatic "Mercedes always pays" guarantee. Drive Pilot's operating conditions are narrow: only on mapped highways without intersections, during the day and in good weather conditions; below 40 mph in the US and 60 km/h in Germany. Before the system hands over control 10-second warning; tunnel, roadwork, bad weather, or emergency vehicles on the road trigger the handover. It is even designed to detect emergency vehicles via microphones and cameras and give an advance warning - because while liability rests with the manufacturer, the software is required to comply with the law.

So In Which Situation Does the Bill Fall on the Driver?

Every case looks at concrete facts. The questions to ask are simple: Did the crash happen while Drive Pilot was active? Was the system operating within its approved conditions? Did the driver take over when control was requested? If you activated the system outside its intended conditions or ignored the warnings, liability may shift to you. If a defect emerges in the sensor or software, product liability comes into play instead. The other driver's fault is also examined as usual, separately. So the promise is bright, but the legal reality varies from case to case.

Tesla FSD: Autonomous by Name, Level 2 Legally

Tesla's Full Self-Driving is, despite its name, still a Level 2 driver support system and requires a supervising driver. Unlike Mercedes, Tesla has not assumed liability for unsupervised FSD crashes. This difference shows up brutally in the outcomes of the two brands' crashes.

The most striking example is the fatal Autopilot trial in Miami: The jury found Tesla 33 percent liable for the crash and awarded 243 million dollars in compensation. In June 2026, it was announced that the company had also settled a fatal "Full Self-Driving" pedestrian crash out of court. In the same period, a federal court granted class certification in the class action filed on behalf of FSD owners; one of the plaintiffs had paid 81,790 dollars for the vehicle and options, plus 7,000 dollars for "FSD Capability".

Why so many lawsuits? NHTSA investigations examined numerous crashes in which Autopilot hit stationary emergency vehicles. Tesla argued every time that driver supervision was required; critics pointed to the excessive confidence created by the system's name. US investigations have identified more than 1,300 driver support system crashes, including fatal ones. The answer to whether a lawsuit can be filed is yes; but drivers usually remain partially liable, and no full "Tesla is guilty" verdict comes out.

Robotaxis and Waymo: Who Answers for a Driverless Car?

Things get really interesting when a car with no driver at all crashes. Robotaxi operators like Waymo can be held directly liable if system failure, software error, sensor failure, or negligent vehicle operation contributed to the crash. Having to answer as the developer of the system is conceptually logical - but in the field, things have not played out that cleanly.

In January 2026, a Waymo robotaxi collided with a student in California; an incident in which a robotaxi hit a child in Santa Monica also fueled the liability debate. GM's Cruise case showed a whole different dimension: a pedestrian first hit by another vehicle and then dragged by the autonomous vehicle brought the question of how artificial intelligence reacts in emergencies and who will be accountable to everyone's agenda.

Which State, Which Chemistry-Based Lawsuit Gets Filed?

When US state law gets involved, the picture fragments. In Texas, accident claims can be directed at the system operator, or at the manufacturer or another negligent driver under product liability; moreover, if the share of fault exceeds 50% no compensation can be recovered. In Florida, in accidents caused by system failure, software defects, or company negligence, a lawsuit can be filed directly against the company. Whether Tesla, Waymo, and Zoox receive different treatment even within the same state is still a matter of debate - the roadmap for robotaxi law has yet to be drawn.

The Black Box: The Quietest Witness in Court

Event Data Recorders, or "black boxes," which are found even in older cars, record the speed of the seconds before the collision, seatbelt use, and airbag data. In autonomous vehicles, the picture multiplies: sensor data, camera recordings, and the system's decisions at that moment are also on record. This means that whether the accident was caused by artificial intelligence or the driver largely emerges from this data. Whether Mercedes' 10-second warning chain functioned properly will also be proven precisely from these records.

But there's another "but": black box data is not automatically accepted in court; admissibility is assessed separately. Whether expert testimony on autonomous vehicle data can pass the Daubert and Frye standards is itself a subject of legal debate. Moreover, some states have regulations restricting access to this data; even the constitutional implications of collecting vehicle data are being discussed in academia. The witness sees everything, but the court must first decide whether to hear it.

Insurance and Regulation: Why Don't the Rules Settle Yet?

The legal framework is running behind the technology. Congress has not enacted a specific federal liability law for self-driving cars; the gaps are being patched with state regulations. While some states require at least $100,000 in liability coverage for driverless vehicles, the minimum coverage mandated by states does not exceed $10 million. In California, a new vehicle regulation came into force on July 1, 2026. For now, there is also no additional insurance requirement for the automated driving systems most consumers have access to; the average annual premium for fully autonomous vehicles is calculated at $3,592 .

System	Level	Liability in an Accident
Tesla Full Self-Driving	Level 2	Primarily on the driver; product liability lawsuits can also be filed separately
Mercedes Drive Pilot	Level 3	On the manufacturer while the system is active and under approved conditions
Waymo robotaxi	Driverless	Operator and manufacturer in case of system/software failure

NHTSA Takes the Stage

The federal side isn't sitting idle. The NHTSA granted an exemption for the first occupant-carrying autonomous vehicle; new performance standards enabling Zoox's commercial robotaxi deployment and accelerated testing exemptions were announced. The agency asked states not to regulate AV design - it will set the design standards itself. In July 2026, it sent a letter to ADS developers after detecting a pattern of driverless vehicles obstructing law enforcement and first responder teams; proposed changes to crash prevention rules are also on the agenda. So regulation is on the move, but it isn't settled yet.

Where Do the Promises and the Legal Reality Diverge?

The industry already accepts that liability is shifting from individual drivers to manufacturers and system developers; insurers will need to adapt their products and risk analyses accordingly. But academia doesn't agree across the board: Mark MacCarthy of Brookings argues for preserving the traditional negligence-plus-product-liability standard instead of strict liability; he proposes limiting strict liability only to "inexplicable" accidents that cannot be explained by human error.

So are autonomous vehicles actually safer? Human error accounts for roughly 90-94% of crashes in the US; even though companies tout low crash-per-mile figures, fatal crashes are a reminder that the technology isn't fully reliable yet. And the fragmentation of the legal framework is a separate problem: there is no single federal policy framework, and industry organizations are calling for one. Whether a driverless vehicle would be answerable to the city, the agency, the service provider, or all of them together isn't even clear in the public debate.

The picture as I read it is this: Mercedes took over liability on paper, Tesla is losing or settling lawsuits, Waymo is a party in cases. The technology has let go of the steering wheel, but the law hasn't fully settled into the seat yet. Which do you think will mature first - the systems or the rules? I'd bet on the rules.