Safety is rightfully defined as the number one priority for all regulators who make decisions about vehicle systems like driving assist or autonomous driving on European roads and it’s fair to assume all people involved try their best to accomplish it. But to try one’s best is not always good enough to achieve the best result.
Opposing legal principles between the USA and Europe have a strong influence on regulators, and result in lower safety than is possible on European roads. It is easy to blame regulators like the UNECE for rules that cause danger, but we need to be fair, honest, and realistic about the underlying structural issues that cause them to make sub-optimal, less safe decisions.
If we analyze the available accident data for automated driving vehicles in the USA and compare Europe and the USA, everybody will agree that there is a huge gap between the two, a gap that proves more people in Europe are involved in unnecessary accidents and fatalities than needed. That is a fact, and a scandal.
This article will discuss why this is happening today and what should be done to improve it.
The published accident data from Tesla, to pick one example, show that American regulations of autonomous driving systems compared to European ones are leading to fewer accidents, and if we had similar data from Audi, BMW, Daimler, or other automakers, they would show us the same. It is totally fair to comment that the data can’t be precisely applied to Europe — because the streets, vehicles, and conditions are different in the USA — but that does not eliminate the overall argument.
In the USA, manufacturers are self-regulating their driving assist and autonomous driving systems — be they from Waymo, GM, or Tesla. If an accident happens from an autonomously driving vehicle system, the manufacturer is responsible and risks paying many millions of dollars besides being forced into other painful measures. The burden to test, validate, and approve of a system as safe is primarily the responsibility of the experts of that system, and the best experts are the ones from the manufacturers. Regulators, of course, have to make approvals, but it can’t be expected that they will fully understand all of the details of a self-improving algorithm and its consequences — not nearly as well as the manufacturers can.
That structure in the USA leads to a situation in which innovative solutions are enabled earlier on roads, while in Europe, the Ministries of Transportation and the EU UNECE (a regulator) are required to validate and approve all systems themselves and in depth before they release them for public use. They work with external companies to help them do that, but they are not experts either. In Europe, it is supposed that this is a safer approach, but data proves the opposite is true if we talk about new innovations with automatic driving vehicle systems.
Safety regulations became important in the last century mainly to make sure the vehicle itself is safe — in accidents, with regard to emissions and equipment, and to other traffic participants. However, the vehicle was mainly a piece of hardware until recently. If we talk about driving assist or autonomous driving systems, though, we talk only about software (with a few exceptions, such as sensors, chips, and computers). Until recent years, the software in a vehicle was mainly the entertainment system and solutions that regulated some technical functionality of the engine. Regulator processes that have been defined, improved, and detailed over the decades to control hardware are now applied to software in a completely different and highly complex new technology world. The processes applied are not fit for purpose.
What we ask regulators in Europe to do is similar to asking a toolmaker to program a software solution. We should not be surprised if the result is not sufficient. Europe uses a regulatory method for a purpose it was never designed for. Therefore, we should not wonder that the outcome is not compliant with the rule that safety is the number one priority.
Regulators and Ministries in Europe are responsible and liable for potential accidents, and that’s another reason why they naturally tend to rather deny an innovation — avoiding being called irresponsible — even if data prove on average that new systems would reduce accidents, harm, and fatalities by a large margin. The European approach to legislate and regulate these systems in detail from a central authority is a fundamental issue, because denying an approval keeps the regulator on the safe side even if that causes more accidents and deaths compared to approving it. Therefore, the approval rate is artificially limited, low, and slow, and that is bad.
Many believe that automakers who are lagging behind in driving assist or autonomous driving systems, like German automakers BMW, Audi, Daimler, and Volkswagen, are lobbying against systems like Tesla Autopilot in Europe and do influence regulators, but I have solid reasons to confirm this is not true.
In particular, some of the German automakers have a high interest in getting approvals for systems they invented and want to monetize. BMW, Audi, and Daimler have made public statements around this, but like with Tesla Autopilot, they can’t activate them because the UNECE regulation does not allow them to do so. The fundamental issue is not that other European automakers are trying to hinder these systems from entering our European roads because they can’t compete; the issue is to be found in the legal structure of responsibilities and liabilities.
Close your eyes for a second and assume you are personally responsible for the regulation of driving assists and autonomous driving systems in Europe. Would you approve an innovation that may reduce accidents, as data show, but if you approve it and an accident happens, you would personally be made responsible? Likely not.
Can we expect the regulators to not care about political and legal pressure on them personally if their reputations, careers, and futures are at risk because an accident happened? Likely not.
Do we expect these regulators to put safety first despite all personal consequences and odds and approve something — based on science, data, and facts — regardless of what this means for them? Yes, we do.
The current structure in place for the regulations leads automatically to suboptimal results, and that’s not the fault of the regulators (whether they be the UNECE or the Ministries of Transportation). It’s the result of how the process is designed and how responsibilities and liabilities are defined.
The UNECE (United Nations Economic Commission for Europe) has the objective of regulating safety requirements for driving assists system and autonomous and automated driving. It represents about 53 countries, mostly European Union member states.
The UNECE has installed several working groups that include different organizations and members who partly come from outside of the UNECE to provide expertise. The working party for automated, autonomous, and connected vehicles is called GRVA.
The working group does not use available accident data from Tesla that show the Autopilot system reduces accident rates on average (perhaps by nearly 10 times compared to other cars in the USA). Their argumentation for not using the data is that, in their understanding, the data from the USA does not apply to Europe. The data from Tesla was therefore not used and excluded when making safety regulation decisions.
Tesla Q3 Safety Report
“In the 3rd quarter, we registered one accident for every 4.59 million miles driven in which drivers had Autopilot engaged. (…) By comparison, NHTSA’s most recent data shows that in the United States there is an automobile crash every 479,000 miles.”
Professor Dr. Ferdinant Dudenhöfer, former Chair of Automotive at the University Duisburg from 2008–2020 and now Director at the CAR-Center Automotive Research, has a very clear opinion of the interpretation of the Autopilot data from Tesla for Germany.
The CAR Institute in Duisburg has calculated what it would mean if all vehicles in Germany had been equipped with Tesla Autopilot in 2019. According to their calculation, instead of 281,849 road traffic accidents, only 29,413 accidents would have occurred. In other words, there would only have been around 10% of the accidents that happened, 90% of road accidents could have been avoided.
“This shows the long-term potential of a safety technology such as this (Tesla) Autopilot,” Dudenhöfer said.
Although it’s not realistic that, as calculated by Prof. Dr. Dudenhöfer, all vehicles will have a system like the one from Tesla, every single vehicle that has one can help to reduce accidents. Even if only a fraction of the Tesla vehicle fleet in Europe today could use the system as its regulated in the USA, accidents would be prevented and injuries and lives would have been saved. Every life and accident counts, and I believe this is not taken into account today.
The Tesla Autopilot version that I had the pleasure of driving in California in September 2020 is not restricted by local authorities as we experience it in Europe. The UNECE claims the regulations put in place for Europe have been decided for safety reasons but the data prove that the limitations forced on the system cause the opposite — less safety. Critical components of the system are reduced in Europe, which does not add safety but adds risks instead.
For everybody who never drove a Tesla on Autopilot in Europe, it can be said the system is regulated in such a way that it drives like an old, uncertain man or woman with extremely slow reaction time, bad eyesight, and reduced oversight. We have all sooner or later experienced such drivers on the road and know they cause danger and need to be treated with extra caution.
That is true in particular on the German Autobahn, where people can drive fast. If someone approaches you in the left lane going 200 km/h (124 mph) from behind, you don’t want to be a driver with bad eyesight and slow reaction time, and you don’t want a semi-autonomous system that could react fast and agile but isn’t allowed to do so because regulators didn’t approve it due to liability and personal reasons. I do not believe we can and should expect these regulators to decide on systems in the way we do today, because it is simply not fair.
Tesla Autopilot in the USA performs lane changes like an experienced driver with above-average eyesight, incredibly fast and safe reaction ability, and perfect timing. The UNECE regulators instead mandated slow reactions, slow driving decisions, and strange maneuvers, including bad eyesight, for Europe — and that is irresponsible.
Imagine you as a driver would switch back from an in-process effort to overtake a car on the Autobahn because the time you defined for it is over regardless of the situation you are in. As a result, the vehicle drives back in the previous lane despite having completed the lane change move almost completely, causing a lot of confusion to everybody around you. That’s what the UNECE regulator mandates Tesla Autopilot do. Confusion often leads to accidents. No one would teach a student to drive like this. Therefore, we must in the name of safety ask why the regulation mandates an unsafe process today.
If it is claimed to be safer, then I would like to see an analysis and data that prove it. A fact-based discussion rooted in science uses data for good reason.
Another argument is that to compare the data from Tesla against all other vehicles makes no sense because “all other” includes many old, not-that-safe vehicles. That’s without a doubt true, but would only explain a portion of additional accidents happening.
It has never been a good idea to make the perfect the enemy of the good.
If we assume for the sake of argument that the claim from the UNECE is justified, that the USA data can’t be applied in Europe, what would hinder the regulators from using European accident data instead? Tesla can enable, for instance, 1,000 Autopilot test drivers in Europe to test the US version on European streets and collect the same data, and UNECE can use the data as a basis for a fact-based decision on what is safe or not safe for Europe. I volunteer to collect data as a Tesla driver herewith.
Every other automaker can do the same to prove the safety of their systems. This is not about Tesla and the EU regulators, but about how decisions are made for all our safety. Why are we all put at risk because the UNECE declines to work with worldwide scientifically agreed methods?
In addition to the above, the democratic processes within the regulatory bodies are unclear for European citizens, and not at all transparent. It’s not known how they are supervised or controlled. The current 6-month meeting frequency is badly suited for rapid development, as we are experiencing in driving assist and autonomous driving systems today.
The meetings are not open to the public to attend and it is unclear how decisions are made. Supervising bodies, regulations, and processes are unknown, and it would be helpful for everybody to establish transparent and frequent communication that allows everybody who wants to understand how decisions are made to follow along.
Many people in Europe are speculating about the UNECE’s decisions, and that is not helpful for anybody. While I am certain they have the best intentions, transparency, control, and democratic processes are fundamental pillars in Europe and it is essential to involve citizens to understand who today feels excluded and kept in the dark. I urge the decision-makers to start open communication and explanation with European citizens to avoid wild speculation, frustration, and sometimes even anger. That’s helpful for them, us, and the acceptance of decisions. I am certain there is nothing to hide here, but openness creates trust and credibility, and that is urgently needed.
Because the media and the public have no access to meetings, it is impossible to influence how decisions have been made today and on what basis and if they reflect the interest of the represented countries, citizens, and safety requirements. Every court case — well, with very few exceptions — is for good reason open for the public to attend, because every citizen has the right to understand under what conditions legal decisions are made.
The German parliament in Berlin has for good reason a transparent visitor balcony where every citizen can join any debate and receive assurance that transparency is a key element of a working democracy in my home country. What applies to laws applies to regulations, and that begs the question of why the UNECE meetings are not open for the public to attend.
Software changes all the time and should not be confused with the days when vehicles changed their hardware only slowly. The UNECE meeting only 2 times a year itself is inadequate. The rapid improvements that are happening with autonomous driving systems (e.g., Tesla FSD Beta) show us that at least a quarterly meeting would be required to decide, for safety reasons, what should be implemented rapidly in Europe or can be delayed.
Would we want the innovation of a safety belt not to be not looked at by regulators for another 6 months because that’s their meeting schedule and they don’t have time? Harm can be avoided if the meeting frequency of the UNECE is synchronized with the pace of innovation of software.
Every life and accident counts but to wait many months and keep a potentially outdated regulation in place because the next meeting is in half a year is itself a safety risk. A timely decision process is critical to comply with safety on our streets. To be fair, for many years, the systems have not developed very fast. Therefore, the infrequent meeting schedule could be justified for the past. My request to the regulators, though, is to acknowledge that things change much more quickly now and UNECE should adjust accordingly.
UNECE oddly poor transparency appears like an attempt to control information, which does not bold well for a democratic organization from the United Nations. That is certainly not intended and therefore should be corrected as soon as possible. I urge the regulators to change in order to prevent people who are interested well informed from spreading wild conspiracy theories. We already have enough baseless claims circulating about the Covid-19 pandemic, democratic processes, and politics in general. We do not need more of that. I would appreciate it if UNECE, the Transport Ministries, and the EU Commission would help us all to avoid a grassroots initiative born out of poor information, uncertainty, frustration, and anger, which could undermine their credibility of autonomous driving vehicle regulation further, and could even harm the EU and UN more broadly.
As a European citizen, I suggest that a regulatory structure be put in place that helps UNECE, the Ministries of Transport of all involved countries, and the EU Commission make decisions that put data, science, and the safety of all citizens as their first priority. Good decisions in the interest of all our safety will only happen if the regulating authorities can put liabilities at least partly in the responsibility of automakers.
While many in Europe believe the following is risky, it is time to realize that the complexity of self-learning algorithms within autonomous driving systems does not allow someone who is not an expert to conclude what is more or less safe. It is unrealistic to assume a central regulator like UNECE will be able to decide on detailed safety regulations regarding software systems they are not experts on for all automakers. The innovation of these systems is accelerating very fast, which does not even allow the regulators to start building the required expertise. They will always be behind. And constantly delayed approvals will mean more and more accidents and fatalities.
The accelerating pace of innovation, with increasingly less involvement of humans in charge of programming, requires a new regulatory approach. It is time to put the burden of liability and responsibility at least partly on the automakers.
It would be best for all parties, but in particular for our safety, to install an approach that is somewhere between the US approach (that puts all responsibility on automakers) and the EU approach (that puts all responsibility on regulators like UNECE or Ministries of Transport of different countries).
Such an approach is not new, but already in place today, where, for instance, automakers self-regulate many aspects of a vehicle and its safety — be it with crash tests or other measures. Regulators can always interfere and call back models if they conclude them to not be safe enough. It is the best approach for all parties to install a similar approach in a world where software improves very fast and central regulator bodies are, for clear reasons, not the experts on every little but potentially important detail of the software.
A path that combines both approaches is feasible and should be installed as fast as possible. It will allow us to improve the speed of implementation and simultaneously our safety.
While the next UNECE meeting is planned for mid-December 2020, I don’t expect much improvement with the current regulations in 2021 for all 53 countries that are part of UNECE. That lack of speed will lead to a situation in which the USA and China will soon be far ahead with the implementation of driving-assistance systems and autonomous driving.
That matters a lot, because not only will innovation and industries be lost for Europe, but it also causes additional accidents and lives. Every life lost is one too many, and not acting and changing how regulation is done is causing more harm, not less.
Safety should be the number one priority, but with today’s implemented structures, responsibilities, and liabilities for autonomous systems, in Europe that is not the case.
This article is intended to summarize for the decision-makers within UNECE, the Transport of Ministries, and their authorities why we need change and why we need it now.
Without change, the USA and China will not only be a safer place than Europe; they will also be the home of a newly forming and attractive innovative industry that, according to ARK Invest, will directly account for a market capitalization of $8 billion.
Well paying future jobs and a large business ecosystem will be lost for Europe for a long time and may never come back. Besides safety and jobs, our social system is at risk, which guarantees a social stability we all benefit from today.
We need change and we need it now.
- Allow the best experts to validate driving assist systems — experts from manufacturers.
- Allow the manufacturer to take more responsibility and liability without trading safety.
- Use data to prove or disapprove conclusions and start to collect relevant data for Europe.
- Transparent communication and citizen participation in regulator meetings are essential.
- Adjust regulator meeting frequency to be at least quarterly, or monthly.
- Adjust the balance of responsibilities & liabilities between regulators and manufacturers.