Insurance and Liability Questions Robotaxi Operators Answer Differently

Take the driver out of the car and the fault question doesn't disappear. It just points somewhere else. In conventional driving, a crash investigation is mostly a question about driver conduct: speed, attention, following distance, right-of-way. In a driverless vehicle there's no driver to ask those questions about, so the operating company and the vehicle manufacturer become the natural parties left holding the liability question instead.

How explicitly a jurisdiction answers that question varies a lot. Some have written AV-specific statutes that assign liability directly, spelling out who bears responsibility for what category of incident. Others haven't bothered, choosing instead to let existing product-liability and general negligence law absorb driverless cases the way it already absorbs conventional ones, treating a self-driving system less as a new legal category and more as another product that can be defectively designed or manufactured. Insurance markets have generally followed wherever the liability lands, pricing coverage around whichever party a given framework identifies as the responsible one, which is one reason the same vehicle can carry a different insurance structure depending on which jurisdiction it operates in.

A related complication is evidentiary rather than statutory: proving fault in a driverless crash usually depends on the vehicle's own sensor logs and software decision record, information the operating company controls and doesn't necessarily release voluntarily. A conventional crash investigation can lean on skid marks, witness accounts, and a driver's own statement. A driverless crash investigation instead depends heavily on data that sits inside the company being investigated, which changes the practical dynamics of any liability claim even where the underlying legal framework, statutory or negligence-based, is already settled.

Underwriting the fleet itself is a separate question from underwriting liability for a specific crash. A large company with Alphabet's or Amazon's balance sheet behind it can plausibly self-insure a fleet in a way a smaller, less capitalized robotaxi operator cannot, which means two companies running similar vehicles under the same state's liability framework can still carry very different insurance arrangements behind the scenes, invisible to anyone not looking directly at each company's own filings.

The number of distinct corporate entities involved in a single ride adds another layer on top of the basic fault question. A ride booked through a ride-hailing app might involve one company operating the app and customer relationship, a second company supplying the autonomous-driving software, and a third that manufactured the vehicle itself, an arrangement Uber's shift toward partnering with outside self-driving providers rather than running its own AV unit illustrates well. Untangling which of those parties bears responsibility for a given failure, a software error, a sensor malfunction, or a manufacturing defect, is a more complex exercise than assigning fault when a single company owns the entire stack from software to vehicle to app.

The UK took the first route. Its Automated Vehicles Act sets out a statutory framework for exactly this question rather than leaving it to case-by-case negligence claims, and the specifics of what that framework covers are laid out in this site's briefing on the UK's Automated Vehicles Act.

Germany's approach sits in a different regulatory tradition entirely, shaped by its own vehicle-approval and civil-liability framework rather than a UK-style standalone AV statute. The specifics are covered separately in this site's page on autonomous driving in Germany.

None of this is legal advice, and it shouldn't be read as that. What's worth noticing instead is how differently the same crash would be classified depending on which side of the Channel it happened on: a UK statutory claim in one case, ordinary negligence litigation in the other.