For a century, road traffic liability has rested on a simple question: what did the driver do? Automated vehicles break that question apart. When the software was steering, accelerating and braking, the person in the seat may have been a passenger in law as well as in fact. The disputes that follow will look different, and the evidence that decides them will be different too.
The legal framework in England and Wales
The United Kingdom has moved earlier than most jurisdictions. The Automated and Electric Vehicles Act 2018 made the insurer of an automated vehicle directly liable to victims for damage caused while the vehicle was driving itself, with rights of recovery against manufacturers and others. The Automated Vehicles Act 2024, building on the Law Commissions’ joint report, created the regulatory architecture: an authorisation regime for self-driving features, an “authorised self-driving entity” responsible for how the vehicle drives, and a new category of “user-in-charge” who is not liable for the dynamic driving task while a self-driving feature is engaged.
The policy intent is that the person in the seat should not carry the blame for what the system did. The consequence for disputes is that the real questions move to the system, its operator, its data and its maintenance.
Robotaxis are arriving
Pilots of fully driverless passenger services have been announced for London, with operators including Waymo and Uber, the latter working with the British company Wayve, having set out plans to run trials. Elsewhere, Waymo’s robotaxis have been carrying fare-paying passengers in several United States cities for some years. The vehicles are heavily instrumented. Every trip generates sensor logs, decision records and video that would have been unimaginable in a conventional collision investigation.
Where the disputes will be
Some will be familiar: a pedestrian injured, a cyclist knocked off, two vehicles in contact at a junction. The novelty is in causation and attribution.
- Was the self-driving feature engaged? The answer determines who was driving in law. It is a matter of record in the vehicle’s data, but obtaining and interpreting that record requires engineering literacy.
- Did the system behave as authorised? Automated driving features are authorised for particular conditions: road type, speed, weather, lighting. A collision outside those conditions raises questions about handover to the human, and about what warnings were given and when.
- Did the other road user behave reasonably? Human drivers interacting with automated vehicles do unexpected things. Contributory negligence does not disappear because one vehicle was driverless.
- Who pays, and who recovers from whom? The insurer pays the victim. The argument about manufacturer, software supplier, fleet operator and maintenance contractor then follows, and it is a technical argument.
What aviation already learned
None of this is entirely new. Aviation has lived with highly automated vehicles for decades, and its accident record contains a clear lesson: the dangerous moment is the handover. Autopilots fly for hours; the human is expected to take control in seconds, often with degraded awareness of what the system has been doing. Investigators learned to ask not just “what did the pilot do” but “what did the automation present to the pilot, and when”. Road transport is about to inherit that question, and the discipline of answering it from recorded data rather than recollection.
The evidence changes
Event data recorders, required on new vehicles under UN Regulation 160, already capture the seconds around a collision: speed, braking, steering, throttle, restraint deployment. Automated vehicles record far more: what the sensors saw, what the system classified, what it decided and why. The investigator’s task shifts from reconstructing the collision from marks and damage to interrogating a dataset and asking whether the system’s decisions were reasonable.
That is good news for accuracy and challenging news for the traditional settlement process. Few lawyers, insurers or mediators are equipped to read that evidence, and disputes that nobody in the room understands tend to settle badly or not at all.
Why this matters for dispute resolution
Automated-vehicle disputes will frequently be multi-party, technically dense and commercially sensitive. Manufacturers will not want authorisation questions aired in open court; insurers will want recoveries resolved quickly; injured people will want answers. That combination points towards confidential, expert-informed mediation, with a neutral who can follow the engineering as well as the law. The technology is new. The need for someone in the room who understands both the people and the data is not.