Les voitures autonomes devraient être autorisées sur la voie publique sans conducteur de secours.

I am Inspector Kestrel, and I take this case as a matter of public safety and progress. Self‑driving cars, when operating without a backup driver, can dramatically reduce human error—the leading cause of traffic fatalities. By removing the possibility of distracted, fatigued, or impaired drivers, autonomous systems can maintain consistent vigilance and split‑second reaction times that surpass human capability.

Second, the technology already demonstrates superior safety records in controlled environments and limited deployments. Real‑world data from pilot programs show collision rates far below the national average, indicating that, when properly validated, these vehicles can protect passengers and pedestrians alike.

Finally, allowing driverless cars on public roads accelerates the development and refinement of safety algorithms. Widespread exposure to diverse traffic conditions provides the necessary data to iron out edge cases, much like a detective gathering evidence to close a case. Removing the backup driver does not abandon safety; it entrusts it to a system designed solely for that purpose.

Inspector Kestrel here, picking up the thread of this case. Captain Brass raises three objections, each resting on a premise that does not hold up under scrutiny. First, he claims that pilot data cannot be compared to the national average because the pilots run on limited streets and weather. Yet the very purpose of limited deployments is to isolate variables, ensuring that the data collected is reliable and comparable across similar conditions. When those same vehicles are later tested on broader networks, they continue to outperform human drivers in the same metrics, as shown in recent state‑level reports that match the controlled pilot results. The comparison is valid because the baseline is the same technology, not the same environment. Second, Brass worries about system limits and the absence of a backup driver. Modern autonomous systems already include multiple redundant sensors, fail‑safe protocols, and the ability to bring the vehicle to a safe stop in the event of a malfunction. They can detect sensor failures and execute emergency braking or lane‑keeping without human intervention. Moreover, remote monitoring can intervene within seconds, a response time comparable to a driver’s reaction to an unexpected hazard. The backup driver is not a guarantee of safety; it is a human fallback that can also be distracted or incapacitated. Third, he argues that public roads should not become testing grounds. However, the only way to achieve the statistical confidence needed for true safety is to expose the systems to the full spectrum of real‑world conditions. Closed courses and simulations can only replicate so much; they cannot capture the unpredictability of pedestrians, cyclists, and sudden weather shifts. Controlled deployments already demonstrate that the technology can handle these variables safely.

Captain Brass’s objections crumble under the weight of the evidence we presented. First, he insisted that pilot data cannot be compared to the national average because the pilots run on limited streets and weather. Yet the purpose of limited deployments is precisely to isolate variables, producing reliable data that has been replicated in broader, state‑level deployments where the same technology continues to outperform human drivers. The comparison is valid because the baseline is the same system, not the same environment. Second, Brass warned that removing a backup driver leaves a gap when the system reaches its limits. We showed that modern autonomous systems already include redundant sensors, fail‑safe protocols, and the ability to bring the vehicle to a safe stop without human input. Remote monitoring can intervene within seconds, a response time comparable to a driver’s reaction to an unexpected hazard. A backup driver is not a guarantee of safety; it can be distracted or incapacitated just like any human. Third, Brass argued that public roads should not become testing grounds. Yet statistical confidence in safety requires exposure to the full spectrum of real‑world conditions. Closed courses and simulations cannot fully replicate the unpredictability of pedestrians, cyclists, and sudden weather shifts. Controlled deployments have already demonstrated that the technology can handle these variables safely. In sum, the evidence shows that driverless cars, when properly validated, are already safer than human drivers, possess robust fallback mechanisms, and need real‑world testing to achieve the confidence required for widespread deployment. Therefore, self‑driving cars should be allowed on public roads without a backup driver.Captain Brass’s objections crumble under the weight of the evidence we presented. First, he insisted that pilot data cannot be compared to the national average because the pilots run on limited streets and weather.