Self-driving vehicles (SDVs) offer great benefits for society, but also need to be carefully assessed and regulated before being integrated and used on our roads. In the following pages, we present a possible scenario for SDVs in 2025 in five key sections, concluding with what we can do in 2019 to ensure that we see a desirable future unfold, while avoiding some of the pitfalls outlined in the scenario. Our vignette discusses a München native, Adrian, and his use of an SDV in his hometown, portraying his use of the vehicle in 2025. This scenario also illustrates the main drivers and inhibitors that may affect the successful integration and adoption of SDVs between 2019 – 2025. The next section looks at possible ethical, legal, social and economic impacts of SDV use in the year 2025, which provides guidance on ways that these harms could be mitigated, while developing approaches that would accentuate the positive impacts of SDV use. The final section concludes with practical steps that we need to put in place now in order reach a desirable future for SDV use in the future.
In the next section, a vignette portrays a narrative of SDV use in 2025. The vignette serves the purpose to visualise how people will use this technology and how it will be adapted by 2025. This is followed by two sections to highlight potential driving forces behind the usage of SDV and what may be the inhibitors to its adaptation by 2025. SDVs are set to have a huge impact on our lives, so section 4 identifies social, ethical, legal and economic impacts; and how these may materialise by 2025. Following from this section, we review ways that we can mitigate the negative and accentuate the positive impacts of SDVs through policy implementation from 2019-2025. The concluding section will bring us back to the year 2019 in order to establish the steps we need to take towards a desired future and avoidance of an undesired future by 2025. The 2025 future outlined has desirable and undesirable features, which this section will try to address to provide recommendations. All of the scenario sections are written from the perspective of someone in 2025, except the final section on recommendations, which is written from the present.
Before leaping into the future, let us briefly consider the technologies that make SDVs possible.
SDVs gained public recognition through the three DARPA challenges in 2004, 2005, and 2007; which resulted in the establishment of four SDV characteristics: sensing, perception, planning and control. Sensors are used to take raw data measurements, which are transformed by the perception component into usable information. The planning component creates a path based on that information, and the control component contains the actuators to drive the car (based on the planned path, through direct sensing, in order to avoid obstacles). A combination of camera, radar and laser systems are used to retrieve data about the environment. For the position and motion of the car, SDVs are equipped with satellite navigation, inertial and odometry measurements.
LiDAR (Light Detection and Ranging or Laser Imaging Detection and Ranging) lasers are the primary sensors used in environmental perception. LiDAR technology works by rotating a laser sensor, providing several million data points per second, creating detailed maps of the nearby surroundings for detecting static and moving objects. Velodyne LiDAR is the best way to detect and track static and moving objects in urban traffic. By removing all ground points and performing 3D clustering on the remaining points, hypotheses can be made on the motion of objects. Clusters can be classified into categories such as cars, bikes and pedestrians.
Great advancements are being made towards road perception by transforming road shapes and markings from 3D estimates into 2D images. There have been developments in computer vision to interpret traffic lights and signals, but more work is required. Advancements in computer vision, combined with LiDAR technology, has enabled self-driving vehicles to overcome many issues related to poor performance at night, ambient lighting conditions, and bad weather conditions. Developments in electronic mapping have aided the car’s navigation by incorporating geographical characteristics, traffic information, building information, and traffic signs. The satellite navigation systems – GPS, GLONASS, Galileo, and Beidou – have spurred developing of electronic maps.
The navigation process of SDVs can be classified into four sections: route planning, behavioural decision-making, motion planner, and vehicle control. The route planning stage involves selecting a specific route to the destination from digital map data. This is done by representing the road as a directed graph with edge weights corresponding to the cost of riding over a road segments, a suitable route is then found in the road network graph. The behavioural component monitors traffic information and observes the behaviour of other vehicles in order to reach its destination. Adapting to real-world uncertainties, and the intent of other traffic participants, has been one of the biggest challenges for SDVs; with developments in machine learning techniques, such as Gaussian mixture and regression models, enabling better traffic trajectory predictions.
There have been vast improvements in higher machine learning capabilities, with much of the cognitive automation being done with advanced deep learning and neural network-based models, such as recurrent, generative adversarial, and long-short term memory. There have been great developments used in hardware of the past number of years, allowing for faster processing of neural networks. Motion planning determines the best path for a car to take, comfortable for the passenger, while avoiding collision. The trajectory calculated by the motion planning is performed by selecting the appropriate actuator inputs based on the planned movement, and the vehicle control tracks without a feedback controller loop system. Early car-to-car developments use small radio transmitters and receivers on each car to broadcast information about location, speed and direction to other vehicles, to determine safe lane changed and merges.The DARPA Urban Challenges emphasised the importance for SDVs to access each other’s information to effectively map trajectories, share their driving data, and update their digital maps. The early developments of SDVs concentrated on self-contained vehicles, but it was not until they incorporated vehicle-to-vehicle communication (V2V), as well as vehicle-to-infrastructure communication (V2I), did they truly progress/
In 2025, self-driving vehicles (SDVs) are used in different urban areas throughout the world. 38-year-old Software Developer Adrian uses his self-driving car to go to his office in München every morning, which was one of the first places to roll out SDVs. “So far, so good”, explains Hans, who has been using his SDV for over 4 months now. “I am able to work in my car while commuting. When you factor in an hour commute each way, I get back 10 hours of my life that is lost in the commute every week. I sit back with my laptop, while listening to Spotify. It’s great!”. Hans’ Waymo Centauri b is one of the few permitted self-driving car models on the market and has been one of the most widely adopted of these vehicles, so far. The Centauri b is still in the hybridisation stage towards full automation, having both automated, semi-automated, and manual driving possibilities at level 4 automation. Legally, Hans can only drive fully automated within designated areas of München, but for most other places the car must be in semi-automated or manual mode. “It is a nuisance when I have to drive outside München. It takes a while to get used to the wheel again. But I understand that it will take other cities time before they catch up with us,” Hans claims, as the vehicle navigates through his neighbourhood in level 4 automation. His car changes lanes and stops at pedestrian lights, gives way at roundabouts, while allowing him the comfort to catch up on work or just relax and take in the scenery.
So far, SDVs have gained universal integration in only seven cities in the world, but there are hopes that this number will increase dramatically by 2030. Many of the leading car manufacturers and experts estimate that this number will be between 50 – 70 cities by the end of the decade. Some of the most pioneering and revolutionary developments have been coming from Silicon Valley, while the most prolific countries behind SDV development have been the US, South Korea, the UK, Japan, China, and Germany. The US has been the real innovator behind SDVs, with more than 40 cities piloting SDVs as far back as 2017, dwarfing all other countries in comparison. At the start of 2025, there were 100 cities in the US piloting SDVs, and this number is set to increase dramatically by 2030.
Hans has reaped the benefits of autonomous driving, but only after he passed his SDV driving test. In addition, cities integrating SDVs must also be authorised with the National Self-Driving Vehicle Transportation Board (NSDVTB) and the vehicle owner must be registered with the Department of Self-Driving Vehicles Authority (DSDVA). The vehicle itself must pass strict manufacturing standards before being allowed on the market. Outside of these designated areas, cars must function at level 3 capacity – limited automation. The car senses when conditions require the driver to retake control and provides a sufficient transition time for the driver to do so. Some SDV companies wanted to skip this stage, but the limitations of technological organisation, the interaction with manual drivers, and the lack of infrastructure to accommodate this move have been too problematic. In areas where there are mixed drivers (automation and non-automation), SDVs must have a level 3 option for legal reasons. One of the main reasons behind these laws is to ensure safety, which has been one of the main driving forces behind the development of SDVs in the first place.
Approximately 90 percent of crashes are the result of mistakes by the driver and while road deaths have been decreasing, they were as high as 1.4 million in 2015. Over the past ten years, safety has been one of the strongest drivers for the implementation of SDVs, but we have yet to reap their true benefits because of their relatively low implementation. There is an ambitious goal to have zero automobile-related deaths in the United States by 2050, which may be feasible if they are successfully adopted nationally. As far back as 2017, there have been studies to show that deploying SDVs when they are only marginally safer than humans (say, 10%), it would still have a dramatic impact on reducing road deaths. Policymakers around the world have largely indicated that waiting for SDVs to be far safer than humans (say, 75 – 95%) is not an option because of how long it would take to reach that stage.
One of the main drivers for SDVs has been that they would allow a greater diversity of people to drive, such as the blind, and some of the elderly and disabled population. They may also offer people the ability to work, sleep, read, eat, or watch TV, while driving. Because of the limited use of SDVs, we are yet to see a huge change in road efficiency and reduced traffic jams, which will require a sophisticated and intelligent transportation management systems to accommodate them. In 2014, the American Trucking Association (ATA) predicted that there would be a huge shortage of truck drivers, which would necessitate the development of self-driving trucks. Their prediction of 175,000 drivers by 2024 actually came up short of the reported 215,000-figure taken in November 2024. SDV trucks have also shown promise to reduce carbon emissions through more fuel-efficient driving.
In the cities where SDVs have been integrated, there has been an increase in public transport and car-sharing because of the novelty of being in a SDV, thus reducing overall carbon emissions, in addition to many SDVs being electrically powered. There has been a huge demand for more environmentally-sustainable vehicles since the Kyoto and Paris climate agreements. Cities view electric SDVs as one way to meet their EU carbon emission requirements. However, there is still a concern that there will be an intersection between more commuting as a result of SDV convenience, leading to overall increased car usage. Since 2023, there have been a number of auto manufacturers testing single-user SDVs to bring people from their homes to public SDV buses, which would further reduce our environmental impact, while reducing costs for citizens.
While the price of SDVs is reducing every year, they are still more expensive than non-automated cars. Some people have proposed that SDVs could be shared in order to reduce costs, so that they do not sit idle in people’s garages or parking lots and are be used throughout the day. In addition, fuel costs for SDVs have been lower because of greater fuel-efficiency and when they reach widespread level 4 integration and safety is improved, it will reduce the necessity for airbags and steering wheels. The whole design will change because of narrower, smaller and more economically viable vehicles. Between the 2020 – 2025 period, a large number of new non-traditional players, such as ICT and data analytics companies, have emerged in the SDV automotive market. Many of the smaller automotive companies view SDVs as a threat because they cannot put the same kind of investments into developing these technologies as their larger automotive counterparts, which will lead to many foreclosing in the coming years as a result of market pressures.
Over the years, some have stated that the SDV market is supply-driven and many people do not want to use them. However, SDVs have witnessed development as a result of the need to transport goods, and businesses also view SDVs as another opportunity through the data retrieved from the vehicles. Auto manufacturers have been hugely competitive in the race to develop SDVs, bringing global success and prestige to their company. Companies have been extensively patenting their cars, products, and services to lock customers into their brand. However, the notion of automotive branding has been changing over the past few years, with a shift from luxury, status and appearance, towards efficiency, safety, and functionality.
Efficiency and productivity drivers
As a result of greater driving efficiency, SDVs opened up the possibility of reducing traffic jams and congestion, identifying better routes to take, driving more sustainably, and a reduction of crashes holding up traffic flow. Despite SDVs being heralded as a way where people can get extra sleeping or relaxation time on their commutes to work, some businesses view them as holding the possibility of cutting out needless ‘driving time’, so their staff can work while in the vehicle.
Greater SDV driving efficiency is witnessing a reduction in lane size and quantity of lanes in areas restricted to level 4 automation. Car-sharing has been increasing in these cities, which will eventually mean less public investment in parking lots because cars will be used throughout the day. While the roll-out of SDVs is still new, cities will eventually witness a reduction in healthcare spending because of fewer automobile accidents. SDVs have been reducing the need to live in urban areas for work because people are able to commute from farther away without the strains of previous commutes, relieving resource strain on these areas.
Barriers and Inhibitors
Safety and security barriers
Many different safety issues slowed down development of SDVs. For example, the motion sickness of passengers is an issue SDV developers have been trying to solve for the past decade. Overall, the safety of automated vehicles has been a primary concern amongst road-users, especially following some of the highly-publicised deaths, such as the Tesla Model S in 2016. People have found it difficult to put their safety in the hands of an autonomous machine for fear of technical or systems failures, malfunctions, or just general unreliability of these new vehicles.
While crashes with SDVs have decreased over the past few years, they are still more risk-prone in terms of accidents per mile driven than driver-controlled vehicles. Even going back as far as 2017, figures indicated accident rates for every 48,000 miles driven for SDVs, compared to every 2.08 million miles driven for non-autonomous cars. Mode transitions has raised additional safety issues, such as distraction, loss of situational awareness, and high workload during take-over. All of these factors have proven to be inhibitors to the successful development of SDVs and are issues that are constantly being tested and rectified. Many people have also been worried about the security risk of SDVs, such as hacking, manipulation and malicious activity.
There have been many technological barriers to SDV development, including issues around data security, vehicle security, hacking and cyber-security. Initially, steering systems had built-in processes to determine abnormal instructions, but after a few minor concerns relating to compromised commands, SDVs were implemented with a ‘master computer’ that takes control and brings to vehicle to a safe stopping position in the case of suspicious activity. Auto manufacturers have been pressured to provide increased AI transparency, which has inhibited the speed of development, as have the challenges of ensuring sufficient software and hardware updates. Locations need to have 5G technology access, which has been a limiting factor to SDV integration in many places. Vehicles request relevant information about their current position from the cloud, overcoming the limitations of sensor-based information. Both automotive and ICT companies have also had to invest heavily into their frequency communication infrastructure as there was an unwillingness by governments to finance these systems at the speed required to facilitate SDV integration.
Since SDVs were first developed, there has been a difficulty to establish standardisation between companies and countries. It has been challenging to develop protocols, with some claiming that regulation has been too stringent, halting progress, while others have stated that it has not been stringent enough. Governments have found it difficult to strike an appropriate balance between the two and there has also been a great deal of diversity with SDV policies globally, ranging from extremely detailed and dense (EU, US, and Japan) to non-existent (Eritrea, North Korea, and Somalia).
Economic and geographic barriers
One barrier for SDVs adoption has been their cost and the infrastructure required to facilitate them. It has been costly to implement policies, procedures, and technical arrangements to accommodate SDVs, so they have largely been adopted by wealthier countries. They have mostly remained untested in many of the world’s poorer countries, which is proving to be a key concern in global SDV and social justice circles. Even within richer nations, there has been a wide divergence in acceptance rates of SDVs. For example, willingness-to-pay studies have varied widely amongst nationalities, with many of these divergences remained unchanged since 2017, despite national efforts: ‘Italian participants were most interested in using autonomous vehicles (65 %), followed by the Spanish participants (54 %), the French participants (51 %), the Belgian participants (50 %), the German participants (44 %) and the American participants (32 %)’. Location has played a fundamental role in the acceptance or rejection of SDVs, due to varying local attitudes, reliance on employment in driving professions, and technological capabilities, as well as economic stability of the country. For example, despite there being a greater acceptance rate among Italian and Spanish citizens, the economic instability of both regions over the past decade has inhibited the integration of SDVs.
One of the main inhibitors to the acceptance of SDVs has been a concern around job security. There has been an increased concern in recent years about SDVs replacing taxi drivers, bus drivers, delivery drivers, and anyone dependent on driving as a profession. Many trades unions and organised workforces in these areas have petitioned and protested at the replacement of workers in these sectors. Animosity towards SDVs from these groups has led to isolated incidences of abuse towards SDV taxi managers, destruction of vehicles and protests outside Waymo headquarters in Mountain View.
There has been a lot of negative publicity about SDVs, especially about fatalities, such as Uber’s accident in 2018. There have been many cases of local residents harassing SDV drivers, slashing tyres on vehicles, throwing rocks, and hostility towards them. The media has sometimes been criticised for focusing on many of the negative aspects of SDVs, such as the crashes and fatalities, which has affected public understanding and acceptance of the vehicles. Providing a level of trust amongst the public in relation to crashes, hacks and malfunctions has been one of the greatest challenges for SDVs market integration. Because SDVs are relatively new to the market, it has also been difficult to estimate user acceptance. In many reports, there is an expressed fear that others will have access to your data. Some organisations have even established protocols to ensure that users’ privacy is protected when selling their SDVs.
There has been a difficulty uniting cohesive legal analysis due to national differences on road traffic and transportation. One of these barriers has been determining accountability in cases of accidents. Manufacturers have tried to keep accountability in the hands of the driver, keeping SDVs at level 3 automation. However, this has also prompted some manufacturers to take full responsibility in order to promote trust in their vehicles. The different levels of accountability have led to some confusion in the insurance industry about how to deal with accidents.
Ethical, legal, social and economic impacts
Safety and prevention of harm
In discussions of SDVs, one sometimes hears questions about whether non-automated driving should be banned when we reach a level where SDVs can safely and easily replace non-autonomous driving. While still in the hypothetical stage, once SDVs become prevalent, ‘it seems morally or ethically necessary to prohibit selling and using non-autonomous vehicles’. Because the rollout of SDVs has been so slow, this has not been a pressing question, thus far. Meanwhile, groups such as Humans Against Autonomous Vehicles (HAAV) have strongly opposed SDVs because they are not safe enough to drive and are just “glorified smartphones”.
Another concern is what an SDV should do if there is an unavoidable crash: How should the SDV be programmed and who should determine these priorities. Nobody would buy an SDV if they prioritised the lives of others over the vehicle’s driver and passengers. However, if algorithms aim to protect the driver, they may crash into children or light vehicles, instead of other cars, walls, or lampposts, to protect the driver. Also, if safety is the main concern, they may swerve towards a motorcyclist wearing a helmet, as opposed to one without a helmet, because they would be more likely to survive in a crash. If algorithms target those less at risk, then people may start to take unsafe activities in order to become safe, i.e. cycling without a helmet so that SDVs view you cautiously, thus avoiding collision. Manufacturers want to confine SDVs at level 4 to areas that prohibit non-autonomous vehicles, because the uncertainty of non-autonomous driving is the biggest risk.
Algorithms determine statistical likelihoods that certain groups of people would be more likely to die in a collision. Surveys to identify driving behaviour are inaccurate because some people feel pressured to give more self-sacrificing, altruistic answers, than they would in reality. However, it is naïve to assume that people are generally self-sacrificing in split-second decisions, which has been verified in driving simulations and experiments. Therefore, creating crash algorithms based on social values, or even individual values, is difficult to incorporate within SDV driving algorithms. While there have been guidelines and recommendations, regulation is still not fundamentally clear for SDV programmers.
There has been a concern that in specific life-or-death scenarios, programmed responses may remove control from the human being in specific circumstances. We lose the choice and ability to make split-second decisions that could imperil our lives or those around us, but that we should make these decisions, otherwise it hinders our autonomy. Car manufacturers have been concerned about how to program SDVs in specific scenarios, because these pre-given responses may not correspond to how we would behave in reality. Advocacy groups have claimed that car manufacturers will program the ‘correct’ response or that they may be forced to do so by regulation, which some propose diminishes human autonomy.
There has also been a concern that SDVs are threatening our free will and responsibility, because of the removal of accountability from the individual as a result of overreliance on algorithms and artificial intelligence. Already, in cities where level 4 automation is in place, there has been personal accounts of individuals feeling a loss of control in these vehicles. In other instances, there have been issues relating to lost control because SDVs have been programmed to abide by speed limits and rules of the road. For instance, in California recently, a pregnant woman went into labour and had to be rushed to hospital but was delayed because of the SDV’s speed limit regulation.
There have been many rights-based issues related to SDVs over the past few years, some of which are still only hypothetical, such as: if a car is shared, who owns the car and what rights do you have to it? Policymakers have identified that while SDVs open the possibility for more people to use them than non-autonomous cars, it also poses the challenge of who do you deny the right to use them. As of now, countries are still following non-autonomous driving policies in relation to capacity to drive a car, as most still require level 3 automation. The elderly, blind and disabled are still being disadvantaged, but once SDVs reach widespread level 4 and 5 automation, they will begin benefitting from them.
Insurance and discrimination
There are concerns that SDV data will be used against individuals, and groups of individuals, by insurance companies. Now that cars are able to retrieve a wide array of driving habits, patterns, and behaviours, it means that if insurance companies gain access to this information, which many have already proposed an interest in, insurance could be tailored to meet individuals’ driving performance. While being heralded as a positive move towards providing better insurance premiums to safer drivers, others have proposed that it would infringe on people’s sense of privacy, with the feeling of constantly being monitored in the vehicle. Others have disavowed it because of the imbalance in insurance between manual cars and SDVs – namely, that insurance companies will provide better conditions for SDV drivers who allow their data to be monitored by insurance companies, to the disadvantage of non-SDV drivers.
Privacy has been one of the most fundamental issues concerning the use and implementation of SDVs over the past decade. As a result of the large amounts of data retrieved from SDVs, policymakers need to identify methods to ensure privacy and data security; determine who should have access to this data; how it should be securely stored; and if law enforcement should be allowed to hack an SDV if it is breaking the law. So far, regulators have determined that strong levels of encryption, anonymization and aggregation need to be implemented in order to protect the individual’s personal data. A lot of automobile manufacturers are promoting their DRIC compliant “data remains in car” approach, which attempt to process and integrate data within the car, rather than being transmitted to different service providers or third-parties. This has been recommended since late 2024, but manufacturers have found it technically challenging to abide by.
Data and privacy
SDVs produce huge amounts of data and require large processing capabilities. The massive amounts of data required to operate SDVs have raised privacy concerns about if individuals are identifiable, who has access to this data, and what can be done with it. There has also been debate over whether data acquired from SDVs can be used as legal evidence; for example, if the driver was in control of the car at the time of an accident, could that evidence be used in court to determine liability. Furthermore, concerns have been raised about how long data should be stored; where it should be stored (e.g., on the car’s hard drive, the manufacturer’s cloud platform or an independent cloud platform); who should be granted access to this data; under what conditions; what happens to the owners data when they sell the car; how will the data be protected from being hacked; and who owns this data.
Sensors collect information about the environment, which could be an infringement on bystanders’ privacy. Because car companies are compiling mixed data (both personal and non-personal), it has been a little unclear how they are abiding by the GDPR. In addition, they have also had to incorporate how they were securely and safely protecting privacy in accordance with ePrivacy Regulations (ePR) created to ensure that automotive companies abide by its guidelines. The European Automobile Manufacturers’ Association (ACEA) and the Council of the European Union have been paramount for ensuring that these governments implement the ePR and that those working in the industry follow the recommendations outlined.
People have been fearful that SDVs will be easily hacked because of the abundance of digital infrastructure required for them to work. Criminals could make explicit use of the data that they retrieve, hack the vehicle and get it to perform actions the user is unaware of, unable to undo, or maliciously cause harm to the individual(s) in the car. If cyber-criminals take over a vehicle, they may cause a nuisance with opening and closing windows or other minor grievances or even disable the car’s functionality to read stop signs, maliciously cause the vehicle to crash and harm its passengers or use the SDVs for terrorist purposes to transport remote-controlled bombs. While there is a greater need for transparency from car manufacturers, there is the problem that cars will become more vulnerable as a result. So far, there have been only a few minor issues related to cyber-security, such as the case in London where attackers found weaknesses in the SDVs through crypto malware and were able to extort money from the passengers before releasing control of the vehicle. However, these were isolated incidences and most of the cybersecurity insecurities have been identified by grey-hat hackers before malicious incidences occurred.
There has been a greater emphasis on strengthening counter-measures to avoid these situations. For example, in January 2025, UK police were granted the ability to take over cars that are hacked or under control for malicious purposes. This was done through the use of Decentralised Environmental Notification Messages (DENM), which are messages exchanged between peer-to-peer SDVs and their digital infrastructures. DENM sends messages to the police if there are abnormalities, that indicate that the vehicle has been hacked, and comprises cryptographic signatures, which ensure that the messages being received from the SDV is from a reliable source, through certification and Public Key Infrastructure (PKI) architecture. The certificates are linked with the vehicle at precise times and if the vehicle can be trusted. These anomaly-based detection methods are able to identify a lot of attacks, but miss others, so there have been developments towards remote attestation methods, which check protocols before granting access to services. If there are abnormal issues addressed during this process, that indicate potential hacking, this is relayed to the Police ICT Departments for further testing before intervention.
Many motorists have been concerned about identifying liability in SDV crashes. At levels 0-2, it is very clear that, legally, the driver is completely responsible for the car’s behaviour. SDVs become an issue at levels 3 and 4, because of the uncertainty of who is liable in cases of accidents. It is very important, under law, to identify who is responsible for the vehicle and under what circumstances. So far, some traditional insurance companies have established insurance policies for SDVs, with premiums at the same rate as non-autonomous vehicles, unless the driver grants them access to their SDV data. SDVs raise the issue of who should be held accountable in case of accidents and thus responsible for compensation. Since 2020, some of the main issues relating to SDV liability are:
- Determining accident liability if the driver is allowed to ‘focus attention on tasks other than driving’. For example, if the car is in self-driving mode and the driver is reading, but needs to quickly take control of the wheel, and fail to do so in time, should the driver be held accountable? So far, manufacturers have largely claimed responsibility for crashes at level 4, but at level 3, drivers are not permitted to do other activities that would prevent them from taking control of the wheel.
- Another problem relates to determining liability at the ‘origin of the malfunction’. It has been difficult to identify the point at which an SDV malfunctions to a specific time, making it a challenge to identify liability. Manufacturers have accepted responsibility for most cases of malfunction in recent years.
- If the driver activates the car when they should not have, or they do not take over control when requested, has also been an issue for liability detection. For example, there have been cases where the driver is aware that they will be liable for an accident if they take control of the wheel, so they don’t, thereby placing liability on the SDV’s system.
- Problems arise when there is a critical situation and the driver and car react at the same time. For example, a car driving in front of the SDV brakes and the driver turns the wheel to the right to avoid a collision, while the SDV veers the wheel to the left to avoid the collision. Both actions counteract one another and the car crashes into the back of the vehicle in front. For example, there was a situation in Seoul last year (May 6th, 2024), where this happened to an SDV driver. Luckily, nobody was badly injured, and the manufacturer admitted responsibility after reviewing the driver’s inboard footage.
- Problems have occurred when there is a crash and the driver’s response would have been better than the SDV’s reaction time in the same situation. For example, in one of the Volvo SDV test-runs in December 2021, there was thick fog on the road, impeding visibility. A cow came onto the road, but the SDV was unable to detect it and reacted too late and the car skidded out of control into a bollard nearby. The driver would have seen the cow earlier, with the use of his high-beam fog lights and could have avoided the crash.
Concerns surround SDVs that break the law, when the driver is not required to monitor its actions. It is difficult to determine if the driver is liable because they should have been monitoring the vehicle, or if the manufacturer is responsible because they implemented the SDV functionality that would break the law. So far, in the locations where level 4 vehicles have been integrated, manufacturers state that they are strictly following local laws and rules of the road, so this issue has yet to materialise in reality.
Joy of driving
For many, SDVs take away one of the primary pleasures of vehicles – the joy of driving itself. While for some driving is a necessary ordeal that must be endured, for others, it is a form of pleasure in itself: a sense of control, a form of relaxation, a sense of adventure, and a connectedness with their surroundings, that is being threatened by SDVs. Some groups of driving enthusiasts are setting up affiliations to ensure that SDVs do not engulf their ability to drive in the future, but many say that the death of non-autonomous cars is an inevitability.
Figure 14 The joy of driving
Many years ago, the BRAVE Project was one of the first to highlight that there are different perceptions about SDVs between men and women. Men have had less worry about embracing the technology, while women have been less enthusiastic and more fearful about the safety of SDVs and the difficulty of their use. Male drivers showed a more favourable attitude towards SDVs. Men have been buying SDVs at a greater rate than women, with an approximate 60-40 split in SDV usage. Manufacturers are supporting further research to determine how to increase female acceptability of SDVs.
SDVs hold the potential to reduce inequalities and promote inclusion amongst drivers by allowing certain groups (senior citizens, non-drivers, people with disabilities) access to automobiles that was limited, or unavailable previously. However, because of the low levels of automation, this has not been possible, although many of these groups have indirectly benefitted from the use of SDV ride-hailing.
While SDV car-sharing has not yet materialised because of low levels of automation, they hold the possibility of changing the nature of car ownership in the future. Some propose that SDVs will not remain static in garages or parking lots but will be shared amongst groups of people and used throughout the day, when we get to widespread level 4 and 5 automation. Google’s Waymo has been pioneering SDVs ride-hailing as far back as 2018 and have since introduced preliminary pilots in a number of cities throughout the US. There were a few incidences in 2023, where passengers were not allowed to leave the car because of a glitch in the payment system, but overall, they have been a huge success and are set to expand their ride-hailing globally.
Travel behaviour and demands
It is still unclear if total travel miles increase as a result of ‘travel comfort, convenience, and possibilities for non-drivers to use cars’. So far, the limited integration of SDVs indicates that people travel more often as it eases many of the stresses found compared with traditional driving. In addition, fuel costs have been decreasing in five of the seven cities where level 4 automation has been implemented, because of more efficient driving, while the other two cities showed no change. In the past, it was assumed that insurance costs for SDVs would decline with a lower number of accidents. However, insurance companies are still dubious about the safety of SDVs and have kept insurance costs mostly the same as for non-autonomous vehicles, unless drivers can prove their safe driving through their SDV data. While SDVs initially had a higher number of accidents per mile than traditional cars, this was simply because they were in such early stages of development. Since July 2024, there has not been a fatal accident as a result of SDVs.
SDVs are allowing for closer travel proximity on the road from safer driving, while producing a more efficient traffic flow. With the prospect of sharing SDVs, it may lead to less parking spaces if they are used throughout the day.
What has been happening in some of the cities where SDVs are being used is that drivers are beginning to live further away from the city centres because of the ease of commuting and reduced costs of running their SDV. There is less of a need to live in cities, which has started to see a reduction in urbanisation, allowing for a more evenly spread out population throughout the region. It has started to take some of the strains off amenities and busyness of very congested cities.
There is an uncertainty about whether SDVs are ameliorating or exacerbating congestion levels. So far, people with SDVs have increased their overall travel time because they see it as less of a burden. However, it has been proposed that SDVs will improve efficiency and reduce congestion levels. Early signs indicate that increased efficiency will reduce harmful carbon emissions more than non-autonomous vehicles. SDV developers have been trying to walk the tightrope between ensuring their vehicles are environmentally-sustainable and having economically-affordable vehicles. Some manufacturers have placed a greater emphasis on emission reductions with the foresight that governments are implementing harsher penalties for poorly performing vehicles.
Security and economic impacts
In the past, there were worries that SDVs would lead to job losses for ‘taxi drivers, parking attendants, valet parkers, car mechanics, meter attendants, traffic officers, and potentially bus and freight drivers’. There have also been concerns that there were not enough people to drive trucks in places such as Canada. Many truck manufacturers, such as Mercedes, noticed this trend and capitalised on autonomous trucks, and have been testing level 5 trucks for locations where it is too dangerous or unsuitable for humans to drive, since Autumn 2023. Uber saw that many of its drivers could become unemployed because of SDVs, so they have created computer science, engineering, and maintenance programmes for those interested in upskilling and transitioning professions.
As a result of the large investments and technological capacities of SDV development, we have seen a number of smaller automotive companies beginning to dissolve because they will be unable to compete with these giants going forward. While SDV start-ups flourished in the early infancy stage, the larger players have started outcompeting them with innovation, thus minimising the competitive market of SDV manufacturers.
Luxury vehicle business
Some of the luxury vehicle manufacturers were worried about how SDVs would impact their business models, especially if driving were relegated to a hobby. However, some manufacturers have flourished through this period, with Audi and Mercedes taking leading roles in the SDV market. However, companies such as Ferrari, Lamborghini and Lexus are trying to re-market their vehicles and have begun investing in ‘drive for fun’ initiatives and racing tracks.
SDVs are very expensive, which has limited ownership to rich people. It is difficult for poor people to drive SDVs and may become problematic when it becomes the prevalent form of transportation. There are concerns that the increased safety of SDVs may cause non-SDVs to be seen as unsafe and eventually prohibited from being sold, limiting people to more expensive SDVs.
In the past, it was suggested that SDVs would cause insurance and energy costs to decrease, but we have only witnessed minor changes. While SDVs are hailed as safer, which should have reduced insurance costs, has not materialised in practice.
Tax and ownership
One recent concern is related to the ownership of SDVs and who will be responsible for the taxation, insurance and maintenance of the vehicle, if they are shared. There have been developments in models of car ownership, with some companies, such as Uber, beginning to implement pay-as-you-use ownership models.
There has been a lot of debate over whether governments should maintain existing infrastructure or start implementing a more digitised infrastructure to accommodate for SDVs. So far, SDVs have had to develop to understand human signs, rather than digital signs. Furthermore, there has been a public outcry about governmental investment in SDV infrastructure, with many claiming that it should be partly funded by auto companies. In late 2024, demonstrations in France and Germany called on SDV manufacturers to aid cities pay for SDV infrastructure.
Law enforcement income
There has been a concern, in London and Mountain View, California, that SDVs will impact income generation of law enforcement. With more law-abiding vehicles, there has been a marginal and slow reduction in speeding and illegal parking. While more law-abiding vehicles is obviously a good thing, it still means a lost form of revenue generation by the police.
Electricity and power
While SDVs have been powered by a mix of electric and traditional fossil fuel, there has been a strong emphasis from governments to switch to electric. For example, the UK government stated back in 2018 that more than half of all vehicles on the road should be electric by 2030. SDVs burn less fuel because of more efficient driving. However, in the cities where there have been large rollouts of SDVs, there has been an overall increase in fuel use because of their increased ease of use.
Mitigating the negative and acting on the positive impacts
As far back as 2019, there have been many different actions to mitigate negative impacts, while accentuating the positive impacts, of SDV technology, through national, international and supranational legislation and policy. One of the ways this was achieved was through national standardisation protocols between policy-makers, auto manufacturers, computer scientists, and transportation agencies. Standardisations have been created to ensure sufficient cyber security capabilities for SDVs are developed and implemented; minimum requirements established for the use of sensor technology; safety levels have been incorporated into earlier vehicle regulations to include hardware standardisations; and there have been several layers of enforced testing for different levels of vehicle automation.
National governments have implemented an array of different measurements and regulation to ensure that safety standards are being met. Many countries have heavily invested in their own independent testing, as there were a number of concerns related to scientific bias in manufacturing testing. In doing so, the US, Canada and Japan have created a greater transparency towards SDV regulation. In total, 65 countries have developed their own SDV driving tests and licensing laws, while also enforcing safety regulations on manufacturers to demonstrate that these vehicles are safe to drive prior to being sold.
There have also been strengthened measures to inform the public about SDVs, how they function, and how non-autonomous drivers should interact with them on the road. This has led to a greater public trust, in conjunction with a large increase in media public awareness campaigns from car manufacturers. There has been a greater emphasis placed on the benefits retrieved from the big data of SDVs, but strict procedures and guidelines have been instituted to ensure personal data is anonymised and encrypted in accordance with GDPR, which has been a milestone for privacy protection over the past seven years.
The automobile industry has had to adapt its earlier approach to the design process of their vehicles, with a greater emphasis on responsible innovation and value-sensitive design. The increase in ethical evaluations of SDVs resulted from state-supported initiatives and the establishment of oversight bodies, such as the UK’s Centre for Data Ethics and Innovation, and Singapore’s AI Ethics Council. Manufacturers have also had to increase transparency, while also providing guarantees for the life-span of their vehicles. Free software upgrades are mandatory for a five-year period with all SDVs sold in the US, Canada, the EU, the UK, China, South Korea and Japan.
Where software updates occur on a regular basis, manufacturers have provided extensive guidelines about these requirements. SDVs have a built-in locking system that will prohibit drivers from using the cars unless their systems are updated. The cars also have clear and purpose-driven maintenance notification for drivers. Depending upon the seriousness of the maintenance, vehicles may prohibit drivers from operating. There has also been collaboration and agreement through the SDV Fair Use Initiative (SDVFUI) to ensure fair sharing of intellectual property for increasing safety in vehicles.
Since 2023, it has been evident that incorporating more digital infrastructure on our roads would be beneficial for the successful implementation of SDVs. While we are still in early stages, SDVs could be used more optimally with improved digital and physical infrastructure. Civil society organisations have been decrying the possibility that all citizens will have to pay extra for those making the change to autonomous driving, when they are not the ones benefiting from them. Policymakers have been negotiating with SDV manufacturers and owners about paying higher taxes to fund the infrastructure required to accommodate SDVs.
Steps towards a desired future and avoidance of an undesired future
This scenario has covered a lot of ground and outlined many different issues, risks, and possibilities of SDVs in the year 2025. It is very important to reflect on some of these situations and highlight those that are desirable by 2025, those to be avoided, and how to go about doing this. For example, governments should implement appropriate legislation and regulation on the sale, use and safety of SDVs. While national, international and supranational institutions should be responsible for ensuring that citizens are protected from the over-eagerness of manufacturers to put their vehicles on the road. The SDV automotive industry needs to be well regulated and controlled to ensure the safety of the vehicles through the effective implementation of SDV regulatory institutions.
There needs to be adherence to current regulations for the effective control of data generated, retrieved and used by SDVs. Clear delineations need to be established about what constitutes essential data for the vehicle’s mobility and if this contains personal and private information. There needs to be clear indication that if essential data contains personal or private information, then it should be strongly anonymized, aggregated, and secured, to protect individual’s privacy. If it is non-essential data, then there should be adequate policies to ensure that it is not retrieved or stored as a result of using an SDV, unless explicit and informed consent is given. Governments need to effectively integrate the tenets of the GDPR into the automotive industry to effectively assure citizens that their personal data will be protected if they use SDVs. Automobile manufacturers have the responsibility of identifying the purposes for which the car collects data in order to demonstrate their compliance with data protection law. For instance, there needs to be careful analysis if this data will be used for advertising, customised pricing, or to sell additional products to the car owner, and either ensure the owner is aware of these, and consent to it, or prohibit use of data in this way, altogether.
The data collected within the vehicle may become important for law enforcement officials in situations where SDVs are either hacked or being used for malicious purposes. There are already technical options being developed to ensure that the harm caused in these situations is minimised, police authorities identify issues as soon as possible, while at the same time not infringing upon the privacy of innocent citizens using SDVs. Methods such as DENM, certifications, cryptographic signatures, and attestation methods, require heavy investment by automotive companies and need to be fit for purpose. There needs to be careful statutory regulation, third-party testing, and planning for the security of these technologies.
Police need to be granted permission to identify, access, and control vehicles that have been hacked or hijacked. However, they should only have access in instances where there is a threat to safety and security, not simply for surveillance purposes. There also needs be effective and appropriate peer-to-peer communications with emergency vehicles, regardless of the fact if they are SDVs or not. Fire brigades, ambulance services, police, or governmental cavalcades, may require access to bypass vehicles and SDVs need to be programmed to identify when these vehicles are approaching. Emergency service vehicles need to be equipped with sensors to inform local SDVs of their approach.
One of the undesirable outcomes of SDV implementation in the year 2025 is that the public might have little input into their integration in the market and information about these vehicles. Citizens should be informed about SDV regulation, so it is vital that policymakers receive input and feedback from the public about their needs. Policymakers should consider the needs of all stakeholders, so that policy is created for the public, rather than forced upon them by governments or SDV manufacturers. Policymakers also need to ensure that there is a smooth transition between traditional infrastructure and the digital infrastructure of the future. For the foreseeable future, SDVs will have to use our current road signs, lights and markings to navigate on roads. However, these may eventually be replaced by ‘digital infrastructure’. While this is not likely to transpire by 2025, governments and companies should still begin preparing for this transition.
Society needs a sustainable transportation system, and this may either be exacerbated or improved with the proliferation of SDVs. Policymakers need to take careful steps to ensure that they are not ‘overused’, once they become so convenient to use that people start commuting much further from work. One such possibility is greater investment into SDV public transportation systems to ensure convenience, cost and energy reductions. This may also prevent poorer citizens from being excluded from the transportation system. Furthermore, careful attention must be placed on ensuring more inclusionary SDVs, especially when they reach level 4 and 5. In particular, the elderly, handicapped, and those who cannot drive, may be granted accessibility to SDVs.