Getting AV Regulation Right: Ten Lessons from Shared Mobility
Key Takeaways:
- Much of the current focus on regulating AVs is on safety, but we need to regulate a lot more than that.
- Our experiences with shared mobility, provide valuable lessons in regulation that we should apply to AVs:
- We should anticipate the issues and proactively shape the outcomes we want.
- Regulate rather than ban.
- Avoid perverse regulatory outcomes.
- Give the right level of government the right regulatory role.
- Gathering and leveraging data.
- Providing supporting infrastructure
- Managing public space.
- Avoid taxing for revenue, not outcomes.
- Community service obligations.
- Avoid over-reliance on a single operator.
Next steps:
Get involved. Join a Transport Reform Network working group on developing regulations for AVs. Simply send me an email at russell@transportlc.org with a brief note on where you are based, which timezone and what you bring to the conversation.
Introduction
Cast your mind back to the early 2010s. Uber had just launched. Dockless bikes were appearing on city streets. Electric scooters were about to follow. A wave of shared mobility was arriving, and its advocates made bold promises: fewer privately owned cars, less congestion, cleaner air, seamless connections to public transport, and mobility options for people who had previously been left behind.
More than a decade later, we know how that story unfolded. Ride-hailing did not reduce car ownership; it added vehicles to already congested streets. Rather than complementing public transport, it competed with it, drawing passengers away from buses and undermining the finances of public transport. Dockless bikes and scooters cluttered footpaths, disappeared from cities overnight when operators decided the economics no longer worked, and left regulators scrambling to respond to something they had never properly prepared for. Car share, arguably the shared mobility option with the strongest case for reducing car ownership, has failed to reach its potential in many places, hobbled by taxes and regulations that made it unviable.
Some cities did get it right, or at least got it less wrong. Where regulators were proactive, where data sharing was mandated from the start, where public space was actively managed, and financial incentives were properly aligned, shared mobility has delivered significant benefits. The problem is that we should have and still could do a lot better.
Now, autonomous vehicles are coming. Once again, the promises are bold: safer roads, independent mobility for elderly and disabled people who cannot drive, optimised traffic flow, and a revolution in how people and goods move through our cities and towns. Once again, the risks are being underestimated: increased vehicle miles travelled, hollowed-out public transport and kerb chaos.
The difference this time is that we have been here before. We know that the promises made at launch do not always match the outcomes that follow. And crucially, we have more time than we did with shared mobility; AVs have taken longer to scale than their most enthusiastic advocates predicted, and that delay is an opportunity we cannot afford to waste.
This blog draws ten lessons from the shared mobility that should inform how we regulate autonomous vehicles. They are not lessons about vehicle safety; important as that is, it is already receiving plenty of attention. They are lessons about the harder, slower, less glamorous work of building regulatory frameworks that shape outcomes at a system level: who regulates what, how data is shared, how public space is managed, how services are taxed, how the benefits of new technology reach the people who need them most, and how we protect communities when private operators decide to walk away.
The cities and communities that learn these lessons will be much better off.
Lesson 1 - Anticipating issues and shaping outcomes
The cities that waited for shared mobility to arrive before thinking about how to regulate them locked in outcomes that are now very hard to reverse.
When ride-hailing launched in the early 2010s, the promises were familiar: fewer privately owned cars, seamless first and last-mile connections to public transport, and a more efficient use of urban road space. Many people largely assumed that ride-hailing would simply do what taxis did, only better. We did a poor job of anticipating what actually happened.
We now know that ride-hailing made things worse across system-level metrics. Car ownership did not fall. Congestion increased. Rather than complementing public transport, ride-hailing competed with it directly, with a significant proportion of trips substituting for journeys that would previously have been made by bus or train. Far from being a solution to urban congestion, ride-hailing became part of the problem.
Now we are hearing the same promises again. AVs will reduce car ownership. They will solve the first and last-mile problem. They are simply better taxis. Some of these claims may prove true. But the shared mobility experience gives us strong reasons to be sceptical of confident predictions.
What we can say with confidence is that AVs carry significant downside risks that are already visible and largely being ignored. Increased vehicle miles travelled is perhaps the most serious: Deadheading, the miles driven without a passenger, could dwarf the equivalent problem with human-driven ride-hailing.
Public transport is equally at risk. If AV robotaxis are priced to compete with public transport, which the economics of autonomous operation may eventually make possible, the cannibalisation of public transport we saw with ride-hailing could accelerate substantially. And kerb chaos, already visible in cities where AVs operate in small numbers, will become a serious challenge at scale.
But there are genuine upsides too, and perhaps the most significant has nothing to do with AVs themselves. Because AVs are inherently trackable, every trip generating a digital record, they make distance and congestion-based road pricing technically straightforward in a way it has never been before. Whether we take that opportunity will depend entirely on whether regulators are thinking about system-wide outcomes or simply focused on putting AVs on the road safely.
The question of whether an AV can navigate a junction without causing an accident is important. The question of what thousands, or even millions, of AVs do to the transport system, the public realm, and our societies is currently receiving far too little attention from regulators.
In short, we need to proactively regulate AVs across the full range of system-wide impacts, including congestion, public transport viability, kerb management, and equity of access. The regulations we develop should not be set-and-forget, but should be treated as a starting point, reviewed and updated regularly as we learn more about how AVs are actually affecting our transport systems.
We have been given an unexpected gift: more time than we anticipated. AVs have taken longer to scale than their advocates predicted. That delay is an opportunity. Now we must use it.
Lesson 2 - Imposing Bans
When a new and disruptive technology arrives on city streets, the political pressure to do something visible and decisive can be overwhelming. Banning it is the most visible and decisive thing a regulator can do. It is also, in most cases, the wrong thing to do, and the history of shared mobility gives us considerable evidence for why.
Attempts to ban ride-hailing largely failed because demand was too strong to suppress. The bans drove activity into legal grey areas, reduced regulatory leverage, and ultimately forced negotiated settlements.
AVs are already facing bans, with New York City having moved to restrict AV deployment on its streets. AVs will not stop being developed because New York restricts their deployment. The technology will mature in other cities. The economic pressure to deploy will grow. And when AVs do eventually arrive in New York, as they almost certainly will, the city will have missed out on the benefits of improved road safety and mobility gains for people who cannot drive.
So what should regulators do instead of banning when faced with genuine uncertainty or legitimate public concern?
The cities that regulated shared mobility well did not do so by keeping it out. They did so by engaging with it early, demanding transparency, building institutional capacity, and being willing to adapt their regulations as evidence accumulated.
Lesson 3 - Avoid Perverse Regulatory Outcomes
There is a particular kind of regulatory failure that is harder to see and harder to fix than a simple bad decision. It is the failure that comes from a regulation that sounds entirely reasonable, survives political scrutiny, and is implemented with genuine good intentions and then produces the opposite of what it was designed to achieve. The shared mobility era gave us several clear examples. The AV era is likely to give us more, unless we develop better tools for anticipating them.
The concept that explains this failure is second-order effects: the consequences of consequences. A regulation produces a behavioural change. That behavioural change produces an outcome. If the analysis stops at the first step, does this regulation achieve its immediate objective? The second-order outcome is invisible until it is too late. Some shared mobility regulation stopped at the first step.
Consider the electrification requirements imposed on car-share operators in a number of jurisdictions. The logic was straightforward: car share vehicles drive more miles than average private cars, so electrifying them would deliver significant carbon savings. Requiring operators to use electric vehicles and to fund the charging infrastructure to support them sounded like a sensible, green policy. Politicians could point to it as evidence of climate ambition. Operators could hardly argue against it publicly without appearing to oppose decarbonisation.
What actually happened was more complicated. Electric vehicles cost significantly more to purchase than equivalent petrol or diesel vehicles. Charging infrastructure is expensive to install. These additional costs made car-share economics considerably worse, forcing operators to raise prices, slow fleet expansion, or, in some cases, exit markets entirely. In many places, these pressures contribute to a car share market that significantly underperforms its potential, with fewer vehicles, in fewer locations, serving fewer members than a less restrictive regulatory environment might have produced.
The second-order effect was the one that mattered most: fewer people switched from private car ownership to car share. And the carbon arithmetic of that outcome is brutal. Research consistently shows that each car-share vehicle removes multiple privately owned cars from the road, with correspondingly large reductions in vehicle miles travelled, parking demand, and lifecycle carbon emissions. A regulation that reduced car share uptake in the name of carbon reduction almost certainly increased net carbon emissions by preventing those ownership substitutions. The first-order effect, cleaner car share vehicles, was real. The second-order effect, fewer people giving up their cars, overwhelmed it entirely.
This is not an isolated example. The shared mobility era produced several other regulations where second-order effects undermined or reversed the intended outcomes.
The common thread running through all these failures is the same: the regulation was evaluated against its immediate, first-order objective and found to be sound. The second-order behavioural responses, operators raising prices, reducing fleets, exiting markets, users choosing alternatives or not choosing the service at all, were not modelled, not anticipated, and not monitored once the regulation was in place.
The shared mobility era was not short of good intentions. It was short of second-order thinking. The AV era does not have to repeat that mistake, but avoiding it will require regulators to ask harder questions of themselves, and of the policies they are tempted to adopt because they sound right, before discovering too late that sounding right and working right are not the same thing.
Lesson 4 - Giving the right level of government the right regulatory role.
The shared mobility era gave us several clear illustrations of what happens when the wrong level of government is responsible for regulating a service.
The most instructive example involves car share and the misalignment between where costs fall and where benefits accrue. In the United Kingdom, the benefits of car share, reduced congestion, lower carbon emissions, and fewer privately owned vehicles on the road accrued primarily at the regional and national level. But the costs of supporting car share fell on local government: lost parking revenue when spaces were converted to car share bays, the administrative burden of managing permits and enforcement, and the political pressure from residents who resented losing parking near their homes.
The result was entirely predictable. Local governments, bearing the costs of benefits they did not capture, had reduced incentives to support car share. Many imposed charges that reduced operators' viability, refused to allocate parking bays, or simply failed to engage with the regulatory questions that car share raised. The UK car share market has significantly underperformed its potential as a result.
With AVs, the jurisdictional stakes are considerably higher, and the same patterns of confusion are already emerging.
We are already seeing jurisdictional conflict in the handful of cities where AVs operate at any scale. In the United States, the tension between state and city authority over AV regulation has been particularly visible in California, where the state DMV has authority to grant operating permits while cities have limited power to restrict where those permits apply.
The principle that should guide jurisdictional design for AV regulation is subsidiarity: decisions should be made at the lowest level of government capable of handling them effectively.
The design challenge is not simply allocating responsibilities to the right level, but ensuring that the incentives and the fiscal flows are aligned. If local governments are expected to manage kerb space, enforce operating conditions, and engage communities, they need the resources to do so, which means that some portion of the revenues generated by AV regulation needs to flow to the level of government bearing the costs of regulation, not simply to the level that happens to collect it.
All levels of government will need to play a role. Deciding on who does what, before AVs arrive at scale, is one of the most important tasks in AV policy.
Lesson 5 - Gathering and Leveraging Data
It took a while to agree on the data needed from shared mobility operators. We cannot repeat that mistake with AVs, where the risks of major transport system impacts are even higher. Today, data is still causing issues:
- Different data requirements by jurisdiction. Creating significant increased costs for operators with little benefit.
- Failing to share data. Data should be available for research and planning across jurisdictions.
- Failing to use data. Transport agencies are often swimming in data. However, they often lack the skills to use the data effectively to support decision-making.
AVs generate vastly more data than any previous mobility mode, not just trip origin/destination, but continuous sensor data about road conditions, pedestrian behaviour, near-miss incidents, and infrastructure quality that could transform our understanding of the urban environment.
Jurisdictions with AVs are already requiring a level of data sharing. However, the issues with shared mobility also apply to AVs. We will need to get better at creating data-sharing standards, be able to share data across different levels of government, and develop the skills to use it.
Lesson 6 - Providing Supporting Infrastructure
One of the consistent failures of shared mobility regulation has been the gap between what we asked of private operators and what the public sector was willing to invest in return. We encouraged shared mobility. We regulated shared mobility. But in too many places, we failed to provide the infrastructure that would have made shared mobility work more effectively.
The evidence on cycling infrastructure and bike share is unambiguous. Bike share schemes that launched into hostile road environments, streets without protected lanes, junctions designed around cars, and no safe places to ride consistently underperformed compared to equivalent schemes in cities that had invested in making cycling feel safe and accessible. Cities that built the cycling network first, or alongside the bike share scheme, saw uptake that cities relying on bike share alone could not match.
For AVs, the most immediate infrastructure challenge is pick-up and drop-off. At small numbers, AVs stopping to collect or discharge passengers is a manageable inconvenience. At scale, it becomes a serious problem. Imagine hundreds of commuters all wanting to be dropped off at the same location in the same short window of time, each requiring a brief stop at the kerb, with no designated space, no sequencing system, and no way of managing the interaction between arriving AVs, departing buses, cyclists, and pedestrians trying to reach the entrance. This is the logical endpoint of current trajectories, and it requires infrastructure investment before AVs scale.
Managing this well will require both physical and digital infrastructure. On the physical side, dedicated pick-up and drop-off zones need to be designed into streets. On the digital side, systems that can allocate kerb space dynamically, sequence AV arrivals to prevent bunching, and coordinate between multiple operators and vehicle types will be essential.
Critically, the regulation of AVs should provide the funding mechanisms for this infrastructure. The principle is straightforward: those whose operations create the need for infrastructure investment should contribute to its cost.
Lesson 7 - Managing Public Space
One of the biggest challenges of shared mobility has been managing public space. Scooters and bikes cluttered footpaths, blocking access for wheelchair users, parents with prams, and pedestrians trying to navigate streets that had not been designed to accommodate fleets of dockless vehicles. Car share required dedicated parking spaces, taking kerb space that had previously been available for other uses. Ride-hailing pick-up and drop-off created new friction points at busy locations, vehicles stopping in cycle lanes, blocking bus stops, pulling in and out of traffic in ways that created conflict with other road users. In each case, the problem was not the technology itself. It was the absence of rules, systems, and physical arrangements capable of producing the right outcomes.
Kerb space is finite urban real estate that is often drastically underpriced and poorly allocated. We have never developed a coherent framework for deciding who gets to use it, when, for what purpose, and at what cost. Shared mobility exposed that gap. AVs at scale will make the problem much, much worse.
Walk through the cities where AVs are already operating and the early signs are already visible. Robotaxis blocking cycle paths to complete drop-offs. Delivery robots occupying footpaths, narrowing the space available to pedestrians and creating hazards for those with visual impairments or mobility difficulties. These are minor inconveniences when the numbers are small. They are serious urban management failures waiting to happen when those numbers scale. We need to be preparing for managing public space to cope with AVs at scale now.
Lesson 8 - Avoid Taxing for Revenue, not Outcomes
One mistake that we have made with shared mobility has been allowing the desire to maximise tax revenue to override the goal of achieving better outcomes. The two are not always in conflict. But when they are, outcomes should win, and with shared mobility, they too often do not.
The car share example is instructive, and not just because of the scale of the failure. In many places, local authorities imposed expensive charges for parking permits and dedicated spaces on car share operators. The charges were not unreasonable in isolation, car share vehicles do use public space, and there is a legitimate argument that operators should contribute to its cost. But the charges were set without reference to what car share was actually delivering: fewer privately owned vehicles, reduced congestion, lower carbon emissions, and a genuine alternative to car ownership for people who did not need a car every day but felt they had no choice but to own one. Against those benefits, the permit charges look less like a reasonable cost recovery mechanism and more like a tax on an outcome we should have been trying to encourage.
The contrast with private car parking makes the problem stark. In most places, private car parking is free or heavily underpriced, a hidden subsidy to car ownership that is so deeply embedded in how we think about streets and land use that it is rarely even recognised as a subsidy at all. Car share operators, providing a service that reduces the demand for exactly the parking spaces that are being subsidised elsewhere, were charged commercial rates for the spaces their vehicles needed to function. The price signals pointed in exactly the wrong direction.
In the AV space, there are a range of potential charges with significant revenue streams: road pricing, pick-up and drop-off levies, operational licences, and more. Each of these is a legitimate instrument. None of them is neutral. Every charge on a particular type of AV use is also, implicitly, a subsidy for whatever is not charged, and the distributional consequences of those implicit subsidies need to be thought through carefully before the charging framework is set.
The most important example of this is the relationship between robotaxis and privately owned AVs. If robotaxis are subject to road pricing but privately owned AVs are not, the charging framework will actively favour private ownership over shared use. People and businesses making rational decisions about whether to use a robotaxi or own an AV will face a price signal that makes private ownership look cheaper than it actually is, once the full costs to the transport system are accounted for. The result would be more privately owned AVs, more vehicles on the road, more parking demand, and less of the shared, efficiently utilised fleet that offers the best prospect of AVs making our transport systems better rather than worse. It would be the exact opposite of what transport policy should be trying to achieve, and it would have been produced not by malice or incompetence, but by a charging framework that was designed to raise revenue without thinking carefully enough about what behaviour it was incentivising.
The fundamental principle should be straightforward, even if its application is complex: AV tax and charge frameworks should be evaluated not by how much revenue they raise, but by whether they make the overall transport system better or worse.
Lesson 9 - Community Service Obligations
In rural areas, the inability to drive is not an inconvenience. It is isolation. Residents who can no longer drive, people with disabilities, and young people without licences are heavily restricted from employment, healthcare, and social life when public transport is inadequate and shared mobility does not reach them. AVs represent perhaps the most significant opportunity in decades to address this, but only if we design the regulatory framework to make it happen.
The starting point for that design is an honest reckoning with the economics of shared mobility. Private shared mobility does not make economic sense outside of dense urban cores. This is not a temporary market failure that competition and scale will eventually resolve. It is a structural feature of how shared mobility works: the unit economics depend on high utilisation, and high utilisation depends on density. Below a certain density threshold, the trips are too infrequent, the distances too long, and the revenues too low to support a viable private operation without public support. This is why car share, bike share, and ride-hailing have overwhelmingly concentrated in city centres and inner suburbs, leaving outer suburbs, periurban areas, and rural communities often unserved.
The response in many places has been to impose community service obligations on operators, requiring them to serve areas or provide services that do not make commercial sense as a condition of being permitted to operate in the areas that do. The intention is right. The mechanism is flawed. Mandating service in unviable areas without compensating operators for the cost of that service does not produce good outcomes. It produces operators who comply minimally, price to recover their losses, or find ways to exit the obligation entirely. The people the mandate was supposed to serve end up with a service that is too expensive, too infrequent, or too unreliable to meet their actual mobility needs.
There is a further complication that the shared mobility experience has made clear: even where operators are present, the service does not always work in the way its advocates hope. Car share, for example, does not function well in places without complementary non-car options. A car share scheme in a suburb without good public transport does not work.
These lessons apply directly to AVs. The economics of private shared AV operation are the same as the economics of every other form of private shared mobility: they work in high-density areas, and they do not work in low-density ones. Left to the market, the AV transition will deliver its benefits overwhelmingly to the people who are already best served by existing transport options, while the people with the greatest unmet mobility need, those in rural areas who rely on family and friends because no other option exists, are left waiting for a revolution that does not reach them.
Given that one of the strongest use cases for AVs is precisely in those rural areas, allowing the market to determine the geography of AV deployment is a significant missed opportunity to use a genuinely new technology to address one of the most persistent and damaging inequities in our transport systems.
The answer to this is to implement well-designed community service obligations that optimise the outcomes we are trying to achieve and minimise the unintended consequences. In the case of rural areas, there is likely going to be a need for some form of taxpayer subsidy so that AVs can benefit people who arguably need them the most.
Lesson 10 - Over-Reliance on a Single Operator
When I wrote my blog post on car share, it was written in the wake of a specific event: the withdrawal of the United Kingdom's largest car share operator. A service that members had come to rely on, which had, for many of them, enabled a decision to give up private car ownership, simply ceased to exist. Overnight, the mobility arrangements that those members had built their lives around no longer worked.
This was not an isolated incident. It was a pattern. Private bike share operators have left cities overnight. Scooter companies that had become embedded in daily commutes, which had been actively encouraged by cities as part of their sustainable mobility strategies, have exited markets with little warning when the economics stopped working in their favour. In each case, the service disappeared faster than any alternative could be put in place, and the people who had reorganised their travel around that service were left without recourse.
The pattern reveals something important about the fundamental tension at the heart of private shared mobility provision. Private operators enter markets when conditions suit them. They exit when conditions no longer do. That is entirely rational commercial behaviour. The problem arises when the service those operators provide has become sufficiently embedded in daily life that its sudden absence causes genuine harm. At that point, the commercial logic of the private operator and the public interest in service continuity are pulling in opposite directions, and there is no regulatory mechanism to resolve the tension.
Shared mobility, for all the disruption its sudden exits caused, operated at a level of dependency that limited the scale of the harm.
With AVs, the potential for dependency is of a different order entirely. If AVs successfully deliver on even a portion of what their advocates promise, a genuine, affordable, accessible alternative to private car ownership, a significant number of people will make the rational decision to give up their cars. People in rural areas who currently rely on family and friends for mobility may come to depend on an AV service as their primary means of mobility.
Now imagine what would happen if the operator providing that service suddenly went out of business.
The harm would not be limited to inconvenience. For people who had given up their cars on the basis that an AV service was available, operator exit would mean being stranded without a vehicle and without the time, money, or opportunity to acquire one quickly. The commercial logic that makes operator exit a rational business decision takes no account of any of this, and in the absence of regulation that addresses the continuity risk explicitly, there is nothing to bridge the gap between what is rational for the operator and what is acceptable for the public.
We are going to need regulations that address this risk directly and in advance, not scrambled together after the first major AV operator failure has already caused the harm it could have prevented.
Conclusion
The arrival of autonomous vehicles represents a significant moment in the history of transport. Get the regulatory framework right, and AVs could reduce road deaths, extend mobility to those who have been left behind by existing transport systems, and help us build less congested, more sustainable cities. Get it wrong, and we risk repeating, and potentially amplifying, every mistake we made with shared mobility: more congestion, more vehicles on the road and hollowed-out public transport networks.
The ten lessons in this blog are drawn from real regulatory failures, car-share operators withdrawing from markets, e-scooter companies abandoning cities overnight, and ride-hailing services that made congestion worse. These failures were not inevitable. They were the result of reactive regulation, misaligned incentives, jurisdictional confusion, and a failure to think about the outcomes that we want from shared mobility.
The encouraging news is that, for once, we have time. AVs have taken longer to scale than their advocates predicted, and that delay is a gift we should not squander. The regulatory groundwork, data standards, jurisdictional frameworks, kerb management systems, community service obligation structures, and continuity requirements can and should be laid before AVs arrive at scale, not scrambled together afterwards, when lobbying against restrictions will become more intense.
But that work will not happen on its own. It requires people who understand both transport and regulation to come together across jurisdictions, share what they are learning, and develop frameworks that are robust enough to shape AV deployment rather than simply react to it. It requires regulators who are willing to think about system-wide outcomes, not just vehicle safety. And it requires a recognition that the decisions being made now, about who regulates what, how data is shared, how public space is allocated, how services are taxed and subsidised, will determine the kind of cities and communities we live in for decades to come.
Get involved
If you want to be part of shaping how AVs are regulated, the Transport Reform Network is establishing a working group on AV regulation, bringing together practitioners, policymakers, and researchers from across the world to develop the frameworks we need.
The working group will focus on practical regulatory guidance that jurisdictions can actually use. Whether you are working in local government, national transport policy, urban planning, or the mobility industry, your perspective matters.
To register your interest, simply send an email to russell@transportlc.org with a brief note on where you are based and what you bring to the conversation. We are building a group that spans countries and time zones, because AV regulation is a challenge that no single jurisdiction can solve alone.
The cities and communities that get AV regulation right will look very different from those that don't. Help us make sure more places get it right.
