The World’s First Global Rules for Driverless Cars: A Turning Point for Autonomous Mobility

For more than a decade, autonomous vehicles have been presented as one of the most transformative technologies of the twenty-first century. Technology companies and vehicle manufacturers have repeatedly predicted a future in which cars navigate cities, transport passengers and respond to traffic without human intervention.

Although major progress has been made, the widespread adoption of fully driverless vehicles has developed more slowly than many early forecasts suggested. The challenge has not been limited to sensors, artificial intelligence or computing power. One of the greatest obstacles has been the absence of a common international framework defining how an autonomous vehicle should demonstrate that it is safe.

That situation is now beginning to change.

On 24 June 2026, the United Nations Economic Commission for Europe announced that the World Forum for Harmonization of Vehicle Regulations, known as WP.29, had approved the first global regulatory framework for fully automated driving systems. The decision introduces internationally harmonised safety principles for the development, testing, approval and operation of vehicles capable of performing the complete driving task without continuous human supervision.

The new framework does not mean that driverless cars can immediately operate without restrictions in every country. However, it represents a major step toward creating a shared technical and regulatory foundation for autonomous mobility.

Why Global Rules Were Necessary

Autonomous vehicle developers currently operate within a fragmented regulatory environment. A system accepted for testing or commercial use in one country may face significantly different requirements in another. Definitions of automated driving, approval procedures, reporting obligations and liability rules can vary between jurisdictions.

This fragmentation creates several problems.

Manufacturers may need to design separate testing and certification programmes for different markets. Regulators may struggle to compare the safety performance of competing systems. Consumers may also find it difficult to understand whether a vehicle marketed as “self-driving” requires constant human supervision or can genuinely operate without a driver.

The new United Nations framework aims to establish a common technical language. It defines how manufacturers should manage safety, validate their systems, document risks and monitor vehicles after they enter service.

The regulation received support from major automotive markets, including the European Union, China, the United States, Japan, Canada and the United Kingdom. Such broad participation is important because autonomous vehicles are unlikely to develop successfully under completely different technical standards in every region.

The objective is not to replace national traffic laws. Countries will continue to determine where driverless vehicles may operate, how they should be insured and who is legally responsible after an accident. Instead, the framework creates internationally recognised principles for assessing whether an automated driving system has been designed and validated responsibly.

Driver Assistance Is Not the Same as Autonomous Driving

The distinction between driver assistance and fully automated driving is essential.

Many modern vehicles already include adaptive cruise control, lane-centring systems, automatic emergency braking and assisted parking. Some systems can control the vehicle’s speed and direction for extended periods. However, the human driver generally remains responsible for monitoring the road and intervening when necessary.

A fully automated driving system has a much broader responsibility. Within its approved operating conditions, it performs the complete dynamic driving task. This includes:

- Steering and controlling the vehicle’s direction

- Accelerating and braking

- Observing traffic signs and signals

- Responding to pedestrians, cyclists and other vehicles

- Selecting safe routes and manoeuvres

- Reacting to unexpected road conditions

- Activating indicators and other safety-related functions

When the automated system is responsible for driving, safety can no longer depend on the assumption that a human will immediately correct every error. This transfer of responsibility requires a more comprehensive approval process than the testing applied to conventional vehicles.

The Competent Human Driver as a Safety Benchmark

One of the most significant principles in the new framework is that an automated driving system must demonstrate a safety performance at least comparable to that of a competent and careful human driver.

This requirement appears reasonable, but measuring it is extremely complex.

Human driving performance varies according to experience, age, fatigue, road conditions, local traffic culture and many other factors. Crash statistics may also be recorded differently between countries. Regulators must therefore determine which human-driving data should be used as a meaningful and fair comparison.

Autonomous vehicles have several potential advantages. They do not become tired, distracted or affected by alcohol. Their sensors can observe several directions simultaneously, while their computers may react more quickly than humans in certain situations.

However, automated systems also have limitations. They may struggle with unusual road layouts, damaged lane markings, temporary construction areas, extreme weather or ambiguous instructions from traffic officers. Situations that appear simple to a human may be difficult for a machine that depends on sensors, maps, software models and predefined operating conditions.

For this reason, manufacturers will not be able to prove safety through a single road test. They will need to provide a structured combination of engineering evidence, simulation results, physical tests, real-world observations and risk analyses.

A Safety Management System Covering the Entire Vehicle Life Cycle

Under the new rules, manufacturers must establish an auditable Safety Management System for their automated driving technology.

This means that safety must be managed throughout the entire life cycle of the system, from initial concept development to the vehicle’s final period of operation. It cannot be treated as a one-time inspection conducted shortly before production begins.

The Safety Management System is expected to cover areas such as:

- System architecture and design decisions

- Hazard identification and risk assessment

- Software development and change control

- Supplier and component management

- Verification and validation activities

- Cybersecurity and software updates

- Incident investigation

- Corrective and preventive actions

- Monitoring of vehicles already operating on public roads

Independent authorities must be able to audit this system. Manufacturers will therefore need to demonstrate not only that a vehicle completed certain tests, but also that safety has been integrated into the company’s engineering and decision-making processes.

This requirement may significantly change how autonomous vehicle projects are managed. A company will need to maintain reliable records explaining why particular design choices were made, how risks were assessed and how problems discovered during operation were addressed.

Simulation and Virtual Testing Will Become Central to Approval

Physical testing alone cannot cover every situation that an autonomous vehicle may encounter.

Consider a single urban intersection. The road may be dry, wet, icy or partly covered by snow. Traffic lights may be operating normally or may have failed. A pedestrian may cross legally, appear suddenly from behind another vehicle or behave unpredictably. A cyclist may approach from a blind spot. An emergency vehicle may enter the intersection. One of the autonomous vehicle’s sensors may also be dirty, damaged or temporarily blocked.

The number of possible combinations increases rapidly when different roads, speeds, weather conditions and traffic behaviours are considered.

Simulation therefore plays a critical role in autonomous vehicle development. Manufacturers can use virtual environments to expose an automated driving system to millions of scenarios, including rare and dangerous events that would be difficult or unsafe to reproduce repeatedly on public roads.

However, a simulation produces useful safety evidence only when its accuracy has been demonstrated. The new framework requires manufacturers to establish the credibility of their virtual testing methods.

They must show that virtual models represent real vehicles, sensors, road users and environmental conditions with sufficient accuracy. Simulation results must also be supported by physical testing, proving-ground evaluations and controlled public-road trials.

The future approval process will therefore depend on a combination of several methods rather than a single type of test.

Manufacturers Must Prepare a Comprehensive Safety Case

Another central requirement is the preparation of a formal safety case.

A safety case is a structured body of evidence explaining why a system does not create an unreasonable safety risk within its intended operating conditions. It connects safety claims with technical analyses, test results and supporting documentation.

For an automated driving system, the safety case may include:

- Description of the system’s operating limits

- Analysis of possible failures

- Sensor and software performance evidence

- Simulation and road-test results

- Traffic-law compliance

- Emergency and fallback strategies

- Cybersecurity measures

- Responses to adverse weather conditions

- Procedures for software updates

- Evidence obtained from real-world operation

This approach places the burden of proof on the manufacturer. It is not enough to state that the system is advanced or has completed millions of kilometres. The company must explain why its evidence demonstrates acceptable safety and how the system responds when something goes wrong.

The safety case can then be reviewed by approval authorities and independent experts.

Safety Monitoring Will Continue After Approval

Traditional vehicle approval is often focused on the moment before a vehicle enters production. Autonomous vehicles require a more continuous approach.

Their behaviour can change after production through software updates, revised artificial-intelligence models, new digital maps or modified operating conditions. A vehicle approved today may not operate in exactly the same way several years later.

The new framework therefore requires manufacturers to monitor the performance of automated driving systems after they enter service. Relevant incidents, safety trends and unusual system behaviour must be identified, investigated and reported to the responsible authorities.

This creates a feedback loop between real-world operation and product development.

Data collected from operational vehicles can reveal situations that were not adequately represented during initial testing. Manufacturers can then improve the system, update their risk assessments and introduce corrective measures.

However, software updates must be controlled carefully. An update that improves performance in one scenario could unintentionally create a new problem in another. Each safety-related change must therefore be validated before it is widely released.

Automated Driving Data Recorders and Accident Investigation

Vehicles operating under the new framework must also include a Data Storage System for Automated Driving, commonly known as DSSAD.

This system records information related to the status and operation of the automated driving function. Following an accident or unusual event, investigators may need to determine:

- Whether the automated system was active

- When the system became active or inactive

- Whether control was transferred to a human

- Whether the vehicle requested human intervention

- How the automated system responded before the event

- Whether the system was operating within its approved conditions

This information may become essential for accident investigation, legal responsibility, product improvement and regulatory supervision.

Without reliable data, it can be difficult to distinguish between human error, system failure, improper maintenance, road-infrastructure problems or operation outside the system’s authorised domain.

At the same time, data storage creates privacy and cybersecurity concerns. Authorities and manufacturers will need to determine which data should be recorded, how long it should be retained and who should have access to it.

The Rules Also Address Vehicles Without Traditional Controls

WP.29 did not approve only one new regulation. It also amended approximately ninety existing United Nations vehicle regulations so that they can remain applicable to automated vehicles.

Conventional regulations often assume that a vehicle has a human driver, a steering wheel, pedals and other traditional controls. Purpose-built robotaxis may eventually be designed without some of these components.

The amendments aim to ensure that existing rules concerning areas such as braking, lighting, occupant protection and general vehicle safety can still be applied to innovative driverless vehicle designs.

This is particularly important because future autonomous vehicles may not simply be conventional cars with additional sensors. They may have completely different cabin layouts, seating arrangements and control architectures.

Global Regulation Does Not Mean Immediate Global Permission

The approval of a global framework should not be interpreted as automatic permission for driverless vehicles to operate everywhere.

The regulatory package includes both a new UN Regulation and a Global Technical Regulation. These instruments provide a common technical foundation, but participating countries may still need to incorporate the requirements into national or regional legislation.

Authorities will continue to make decisions concerning:

- Public-road operating permits

- Insurance requirements

- Accident liability

- Remote supervision

- Geographical operating areas

- Passenger safety

- Data protection

- Traffic-law enforcement

- Emergency-service procedures

The new ADS and driver-control assistance regulations are expected to enter into force in January 2027 after the relevant technical procedures are completed.

Their practical implementation will therefore occur gradually and may differ between regions.

Why the Timing Is Important

The adoption of the framework comes as commercial robotaxi deployment is accelerating.

According to the International Energy Agency, the global robotaxi fleet more than doubled during 2025, reaching approximately 8,000 vehicles operating across around twenty cities. Commercial services are currently concentrated mainly in China and the United States, although operations and trials are expanding into other regions.

The IEA identifies companies such as Waymo, Baidu, WeRide and Pony.ai among the main commercial participants. Robotaxi developers are also forming partnerships with ride-hailing platforms and established vehicle manufacturers.

Forecasts for 2035 vary widely, but estimates cited by the IEA range from approximately 700,000 to three million robotaxis worldwide. Most are expected to remain concentrated in a limited number of major cities.

Even the lower estimate would create a major regulatory challenge. Small experimental fleets can be supervised closely, but hundreds of thousands of commercial vehicles require standardised safety processes, transparent reporting and clearly assigned responsibilities.

The new framework therefore arrives at a critical moment: before large-scale deployment makes fragmented regulation even more difficult to manage.

Cybersecurity, Liability and Public Trust Remain Major Challenges

Although the new rules are an important milestone, several difficult questions remain unresolved.

Autonomous vehicles depend on cameras, radar, lidar, satellite positioning, digital maps, communication networks and powerful onboard computers. A failure or cyberattack affecting one part of this ecosystem could influence the vehicle’s behaviour.

Manufacturers must protect the system against unauthorised access while ensuring that essential software updates can be distributed safely.

Liability is another complex issue. When a human driver causes an accident, responsibility is normally assessed through established traffic laws. When an automated system is driving, responsibility may involve several parties, including:

- The vehicle manufacturer

- The automated-driving software developer

- A sensor or component supplier

- The fleet operator

- The maintenance provider

- The remote-support organisation

- The road-infrastructure authority

Determining responsibility will require reliable technical evidence and clear national legislation.

Public trust will also be critical. People may not accept autonomous vehicles simply because they have passed a technical approval process. Regulators and manufacturers will need to communicate performance data, limitations and incident information transparently.

A system that is statistically safer than human drivers may still face public resistance if its decisions are poorly understood or if companies appear unwilling to share information after accidents.

From Technological Promises to Verifiable Evidence

The adoption of the first global rules for fully automated driving represents a transition in the development of autonomous mobility.

For many years, public discussion focused primarily on what self-driving technology might eventually achieve. Companies demonstrated impressive prototypes and made ambitious predictions about future deployment.

The new regulatory era will focus less on promises and more on evidence.

Manufacturers will need to demonstrate that their systems are safe, that their virtual tests are credible, that their engineering processes can be audited and that their vehicles remain safe after entering service.

Autonomous vehicles will no longer be evaluated only as mechanical products. They will be treated as continuously evolving cyber-physical systems whose safety depends on software, data, connectivity, organisational processes and real-world monitoring.

Conclusion

The world’s first global framework for fully automated driving does not mean that steering wheels will disappear immediately or that driverless vehicles will suddenly become legal in every country.

Its importance lies elsewhere.

For the first time, major international markets have agreed on a common foundation for demonstrating the safety of vehicles that can perform the complete driving task without continuous human supervision.

The framework introduces life-cycle safety management, credible simulation, structured safety cases, operational monitoring and automated-driving data storage. It also establishes the principle that an automated driving system should perform at least as safely as a competent human driver.

Many legal, technical and ethical challenges remain. Cybersecurity, liability, privacy, infrastructure and public acceptance will continue to influence the speed of deployment.

Nevertheless, the decision marks an important turning point. Autonomous driving is gradually moving beyond experimental demonstrations and entering an era defined by international standards, independent assessment and continuous accountability.

The central question is no longer simply whether a vehicle can drive itself.

The more important question is whether it can prove—through transparent, verifiable and internationally accepted evidence—that it can do so safely.