Robotaxis will be on London roads by end of this year. As part of the roll-out, Transport for London will have a key role to play and as a result the London Assembly’s Transport Committee has rightly launched an investigation on Autonomous Passenger Vehicles (or robotaxis) in June this year. The Committee is chaired by Assembly Member Caroline Russell.
Along with public hearings and a call for evidence, the committee aims to explore whether and how APVs could be licensed for commercial operations, and what role the Mayor and TfL should play. It also seeks to understand impacts on taxi and private hire employment, road user interactions, and alignment with Mayor's Transport Strategy targets like Vision Zero.
When it comes to policy making or addressing policymakers’ concerns, I prefer to engage in a conversation to provide the underpinning evidence and facts to enable them to shape their positions. As an academic, I feel it is my duty.
Over three sessions of the Transport Committee on the topic, various themes emerged as areas that the committee had concerns. This was brought to limelight especially in the second session where I was giving evidence along with developers, operations, disability advocates and campaigners.
Understanding safeguards already in place
While the committee’s safety concerns regarding robotaxis and their interaction with pedestrians and cyclists are valid, extensive global work has established strong regulatory confidence in the technology. Following seven years of international collaboration co-chaired by the UK, the United Nations World Forum for Harmonization of Vehicle Regulations adopted UN Regulation 185. This groundbreaking framework creates unified safety standards across major automotive markets, including the UK, US, China, EU, Japan, and Canada. Under this global agreement, any autonomous vehicle approved for deployment must strictly demonstrate its ability to navigate safely alongside vulnerable road users.
Data sharing is not a choice but mandatory by law
When trying to regulate or introduce a new technology, it is like going into the unknown. We can only regulate and make things safe on things that we know of, or we can predict will happen.
An important part of the adopted UN regulation is about learning from autonomous vehicle deployment itself. The UN regulation introduces a concept called In-Service Monitoring and Reporting (ISMR), which is a requirement for all developers who are bringing these robotaxis into any city in the UK or across the world to be able to report as per the agreed data requirements at the UN. These data requirements have been agreed by regulators, by industry, and different advocacy bodies.
While the assembly members were concerned that data will not be shared by the developers or the operators, their concern has already been taken care by the UN regulation. If a developer/operator doesn’t want to share the data (as per the UN regulation), then they are not allowed to operate their vehicles. It is very simple.
Training and testing data
Assembly members rightly raised concerns that AI models driving robotaxis might fail to recognise all pedestrians and cyclists equally due to dataset bias. However, developers and regulators have directly addressed this issue through UN Regulation 185 and the UK’s Automated Vehicles Act. These regulatory frameworks mandate that AI training and testing datasets maintain strict representativeness and completeness regarding local road actors.
For deployments in London, datasets must explicitly reflect specific local nuances, including local cycling behaviours, clothing styles, and hand signals. Operators cannot simply import safety claims validated elsewhere; gathering real-world data on London's streets is a non-negotiable requirement for approval.
The UK’s Vehicle Certification Agency will rigorously audit these datasets and challenge developers' claims before granting operational permission. By enforcing localised data collection and thorough oversight, regulators ensure that detection disproportionality is systematically eliminated before any autonomous vehicles are approved to transport passengers.
Passengers will not be alone
In a robotaxi, the passengers will be alone inside the vehicle. The assembly members were wary of how emergency situations will be handled and for the well-being of the passengers.
An often misunderstood or unknown aspect of an autonomous vehicle service (robotaxi) deployment is that no fully autonomous vehicle service, in practical terms, can exist without some level of remote support. Almost everybody who I have spoken to in terms of developers have some level of remote support. It could be a customer call centre where you can press a button on the vehicle, and you speak to somebody immediately on the other side.
This is not an afterthought. It is in-built into the service design and agreed globally. I have had the privilege of co-leading an international standard, ISO 26361 (together with Nissan in Japan) on what are the requirements for this interaction between people inside the vehicle with a remote centre.
Accessibility as a design choice, not an afterthought
When it comes to accessibility and inclusivity, there are two different aspects to it. One is the vehicle design, which is in most cases not the responsibility of the technology developer, because most technology developers tend to just buy a vehicle off the shelf. It could be a Jaguar I-PACE (for Waymo), it could be a Ford Mustang Mach-E (Wayve). Making the vehicle accessible is the responsibility of a vehicle manufacturer. For accessibility, we need to focus on both the vehicle, as well as the service design (a less focussed area).
As the Director of Partners for Automated Vehicle Education UK (PAVE UK), we have just published a six principle charter for responsible Autonomous Vehicle deployment. It has been co-created with signatories including Nissan, Uber and Wayve, and accessibility is one of the six principles. Others being safety, public benefit, local partnership, transparency and responsible integration.
Do we know it all?
Definitely not. While immediate full-scale deployment is unwise, London is on the right track for a safe, gradual rollout through the Automated Passenger Scheme pilot. Since the UK’s full autonomous vehicle regulation will not take effect until late 2027, this pilot scheme allows regulators and operators to learn from real-world operations and uncover unforeseen challenges. Paired with robust data-sharing mechanisms, this approach ensures ongoing safety. Skepticism is natural, so continuous dialogue remains essential, and the upcoming Transport Committee report will hopefully address and allay key assembly concerns.
