Autonomous vehicles are not a technical problem

In France, the deployment of autonomous vehicles is identified as a national priority[1]. Across the world, major technology companies and automotive actors are investing heavily, racing to bring these systems to market.

At this stage, the question is often framed in relatively simple terms: how do we make autonomous vehicles safe, reliable, and acceptable? This framing is part of the problem.

What is really being optimized

Autonomous vehicles are not a single innovation. They are the convergence of multiple systems: artificial intelligence, sensors, infrastructure, data flows, regulatory frameworks, business models.

And within this system, choices are constantly being made.

What level of safety is “acceptable”? What trade-offs between cost and redundancy? What data is collected, and for what purpose? What infrastructures are required, and who pays for them? Who gets access to these services, and under what conditions?

These questions have very concrete implications.

A system optimized for dense urban areas does not address rural mobility. A model based on high upfront costs and complex infrastructures limits accessibility. Replacing drivers with automated systems raises immediate questions about employment and transitions. Collecting large volumes of mobility data creates dependencies and risks that are not neutral.

Yet these elements are often treated as secondary constraints, rather than as central design choices.

When I was invited to speak about responsible innovation at Alphabet's X Development (ex- Google X Labs) - deeply involved in autonomous vehicle development — what was striking was not the absence of concern for these issues. It was how they were positioned.

The discussion is not whether autonomous vehicles should be responsible. It is which dimensions of responsibility can be integrated without slowing down development, increasing costs beyond acceptable thresholds, or creating too much uncertainty for deployment.

Responsibility is part of the equation, but it does not define it.

What “responsible” actually changes

Much of the discourse around autonomous vehicles assumes that innovation can be balanced: that economic, social, and environmental benefits can be aligned if the technology is well designed.

In practice, these dimensions do not carry the same weight.

Economic performance (growth, competitiveness, return on investment) remains central. It determines where investments go, which use cases are prioritized, and which actors can participate.

Other dimensions (accessibility, inclusion, environmental impact) are addressed, but under conditions. Rural accessibility, for instance, depends on infrastructure investments that are rarely prioritized. Environmental gains are often tied to assumptions about shared mobility that are not guaranteed. Inclusion depends on pricing models that may or may not emerge.

These are not implementation details. They are structural trade-offs. By the time questions of responsibility are formally addressed — through regulation, ethical guidelines, or design adjustments — many of these trade-offs have already been made.

At that stage, responsibility becomes a matter of correction: improving safety margins, protecting data, adjusting services.

But the trajectory is already set. This is the limitation of the idea that technologies can simply be made “responsible”. It assumes that responsibility can be added after the fact.

In reality, it depends on who is involved early on, which problems are prioritized, and which constraints are considered non-negotiable.

What this example shows

Autonomous vehicles make visible a broader dynamic.

Technological systems are not neutral. They are shaped by industrial strategies, political priorities, available infrastructures, and economic models. These elements define not only what is possible, but what is pursued.

Once a direction is taken, it becomes increasingly difficult to change. Investments are locked in, infrastructures are built, ecosystems align.

The question is therefore not only how to make a given technology more responsible.

It is whether we are able to influence the conditions under which it is developed in the first place, before certain choices become irreversible.

This raises a more concrete and often overlooked issue: who is actually in a position to shape these technologies?

Who is around the table when early design choices are made? Which actors have access to the resources, infrastructures, and partnerships required to participate? Whose needs are considered legitimate, and whose are not even formulated?

Today, these spaces are largely occupied by industrial actors, large technology companies, and public institutions aligned with their priorities. Other actors (local communities, civil society organizations, social entrepreneurs) are rarely involved at the stages where key decisions are taken.

This is where the question of co-creation becomes central. Without early and meaningful involvement of a broader set of actors, the same patterns tend to reproduce themselves: technologies are developed for the contexts, users, and models that are already prioritized.

The question is often framed in simple terms: how do we make autonomous vehicles safe, reliable, and acceptable? But this question hides a more fundamental one.

Who is this “we”?

Is it the companies developing the technology? Public authorities regulating it? Engineers designing the systems? Or the people whose lives, jobs, and environments will be affected by it?

As long as this “we” remains undefined, the framing itself is misleading. It suggests a collective capacity to act, while decisions are in fact concentrated in the hands of a limited set of actors.

This is why the problem is not only technical. It is about who has the power to decide what responsible innovation means, and for whom.

[1] Les Echos, 31/01 2018.