Why current traffic laws cannot handle autonomous vehicle crashes

According to the study, most road traffic laws were written on the assumption that a human being sits behind the wheel, exercises continuous control and can be blamed if something goes wrong. This assumption is already fraying at the edges as advanced driver-assistance features such as lane-keeping, adaptive cruise control and automated parking take over parts of the driving task.

Lawmakers around the world are under growing pressure to decide who pays when an autonomous vehicle crashes, as advanced driverless systems move from test tracks to everyday roads. A new legal analysis warns that many countries still rely on frameworks built for human drivers, even as control shifts to software, sensors and remote operators.

The study, titled Liability for Autonomous Vehicle Torts: Who Should Be Held Responsible? and published in the World Electric Vehicle Journal, sets out a comparative roadmap for how responsibility should evolve from human drivers to manufacturers, operators and system providers as automation increases. The paper analyses leading jurisdictions in Europe, North America and the Asia–Pacific region, and then develops a tiered liability model using China as an in-depth case study.

Rather than treating driverless technology as a single legal problem, the authors argue that liability must be matched to specific levels of automation and real-world operating scenarios. Their central warning is that clinging to traditional, driver-centred rules in an era of algorithmic control risks unfair outcomes for victims, distorted incentives for industry and persistent uncertainty for insurers and regulators.

Global legal systems struggle to keep pace with automation

According to the study, most road traffic laws were written on the assumption that a human being sits behind the wheel, exercises continuous control and can be blamed if something goes wrong. This assumption is already fraying at the edges as advanced driver-assistance features such as lane-keeping, adaptive cruise control and automated parking take over parts of the driving task.

The authors group existing approaches into four main liability regimes. In the driver liability regime, still dominant for vehicles up to Level 2 automation, the human driver remains the primary legal subject. The system is treated as an aid and faults are pursued through conventional negligence or traffic offences. This model remains common in countries that have yet to adopt dedicated autonomous vehicle laws, and it is often preserved even where new legislation has begun to emerge, particularly for lower levels of automation.

The manufacturer and operator liability regime becomes more prominent as automation reaches Level 3 and above. Here, legal responsibility starts to follow technical control. When the system is in charge, manufacturers, fleet operators or service providers can be held responsible for product defects, software failures or inadequate maintenance. Germany’s requirement for data recorders in automated vehicles and the European Union’s updated product liability directive are highlighted as early examples of this shift.

A third model, the system liability regime, centralizes responsibility in a designated legal entity that oversees the automated driving system as a whole. The United Kingdom’s insurance-centred framework and new regime for authorised self-driving entities illustrates this approach. In such systems, the victim is compensated without having to untangle complex chains of responsibility; the insurer or authorised entity can later seek recovery from manufacturers or other parties.

Finally, many jurisdictions are experimenting with a composite liability regime, particularly where there is no unified autonomous vehicle statute. Under this model, traditional fault-based rules, product liability and mandatory insurance interact on a case-by-case basis. Courts may apportion responsibility between drivers, manufacturers, operators and other parties depending on the facts, while mandatory insurance schemes guarantee at least basic compensation.

The authors argue that each regime has strengths and weaknesses. Driver-centred systems are simple and familiar but become increasingly detached from technological reality as automation grows. Manufacturer-centred and system-centred regimes better reflect control but risk concentrating liability in a small number of firms. Composite systems offer flexibility but can lack clarity and predictability.

A tiered, level-based framework for the autonomous era

To resolve these tensions, the study proposes a structured, level-based model that can be adapted internationally. The guiding principle is straightforward: liability should follow control. Whoever exercises real-world authority over driving tasks and risk management should bear the primary legal burden.

For Levels 0 to 2, where systems are assistive and the human remains responsible for steering, braking and situational awareness, the authors recommend retaining traditional driver liability. This preserves continuity, avoids overcomplicating low-level assistance features and reflects the fact that human vigilance still determines safety outcomes.

The real disruption begins at Level 3, where vehicles can control most aspects of driving within defined conditions but expect the human to retake control when prompted. The study describes this as the “awkward middle” of automation: drivers are encouraged to disengage, yet are still expected to respond quickly in emergencies. In this zone, neither pure driver liability nor pure system liability is adequate.

Instead, the authors call for a shared regime that assigns primary responsibility to manufacturers and system developers in well-defined, infrastructure-supported environments such as highways or designated urban corridors. Where the system fails within its intended operating domain, manufacturers or system providers should bear primary liability. Drivers would carry secondary responsibility if they ignore operating instructions, misuse the system or fail to respond to clear takeover requests.

In more complex or poorly mapped environments, such as dense mixed-traffic urban streets or rural roads, the study suggests a broader composite approach in which operators responsible for remote monitoring, software updates and communications infrastructure share liability with manufacturers and, where appropriate, drivers. This flexibility recognizes that control and knowledge are distributed across multiple actors during the transition phase.

At Levels 4 and 5, where vehicles can execute the full driving task within their operational design domain without human oversight, the study argues that occupants should be treated as passengers rather than drivers. Primary liability should shift firmly to manufacturers, system developers and operators who design, deploy and manage the autonomous systems. A designated authorised entity could be made responsible for regulatory compliance and safety before a crash, while manufacturers and operators would carry strict or near-strict liability afterwards, backed by compulsory insurance and internal recourse mechanisms.

China’s fast-evolving autonomous driving sector is used as a detailed case study for how such a model could be implemented in practice. With extensive test zones, tens of thousands of kilometres of pilot roads and a growing record of software-related recalls, China captures both the promise and the risk of rapid deployment. The authors outline how a level-based liability framework could be integrated into revisions of road traffic laws, urban governance systems and national safety standards, showing how the approach can be translated into concrete statutory language and institutional design.

Insurance, data and governance must evolve alongside liability

Changing liability rules is only one piece of the puzzle. Insurance, data governance and multi-level regulation all need to adjust if autonomous vehicles are to be rolled out safely and at scale.

Traditional motor insurance products are built around the idea of a fallible human driver whose behaviour determines risk. As automation rises and collisions are more likely to result from software defects, sensor failures or data issues, those products become misaligned with the real sources of danger. The authors argue that insurers will need access to operational data from vehicles and systems to underwrite risk properly, which in turn requires clear rules for data sharing, privacy and cybersecurity.

They recommend developing dedicated insurance lines for higher levels of automation, with coverage that reflects the responsibilities of manufacturers and operators as well as any residual obligations of owners or passengers. First-resort compensation models, where insurers or authorised entities pay victims promptly and then pursue recovery from responsible parties, are presented as a promising path to balance victim protection with complex liability chains.

The paper also underlines the importance of national and regional governance structures. As cities and regions experiment with autonomous shuttles, robotaxis and freight pilots, inconsistent rules on liability and data can create uncertainty for companies and the public. The authors call for coordinated frameworks that harmonise national legislation, technical standards and local implementation, backed by shared databases of accidents and incidents.

China again serves as a detailed example of how this coordination challenge plays out, with dozens of municipalities piloting their own autonomous vehicle regulations. The study suggests that other large countries or regions with federal structures will face similar issues, making China’s experience instructive even if its legal system is distinct.

The authors argue that a tiered, scenario-sensitive liability regime can be exported across both civil-law and common-law traditions. Civil-law states can codify level-based responsibility rules in traffic codes or dedicated autonomous driving acts, while common-law jurisdictions can blend statutory guidance with evolving case law and composite liability approaches. In both families, a clear mapping between automation levels, control responsibilities and insurance mechanisms is presented as the foundation for predictable governance.