Privacy Notice for KE:SAI Test Vehicles

KE:SAI is a non-profit research lab that develops self-driving systems and publishes its results openly. For this research, KE:SAI operates marked test vehicles on public roads in Germany and France. This notice explains which data these vehicles record, how we anonymize it, how we use it and which rights you have.

This notice applies to the recording of sensor data by KE:SAI test vehicles in public traffic and, where permitted, on private premises and test sites. The vehicles carry a sticker that refers to this page (kesai.eu/data). Anyone who is near such a vehicle during a recording drive may be recorded.

KE:SAI — Kyutai ELLIS Scalable Autonomous Intelligence gGmbH
Maria-von-Linden-Str. 1
72076 Tübingen
Germany
E-mail: info@kesai.eu

The vehicles are equipped with cameras, LiDAR and radar sensors, a GPS receiver and an inertial measurement unit (IMU). They record camera images, LiDAR and radar measurements, the position and motion of the vehicle and the time of each recording. The vehicles do not have microphones and do not record audio. The camera images can show people, vehicles and license plates in the surroundings of the vehicle.

All camera images are anonymized on the vehicle immediately after they are captured and before they are stored. A detection model finds faces and license plates in every image and covers them with a coarse mosaic pattern that does not allow faces or license plates to be recognized. Only anonymized images are stored. In this notice, "anonymized" means that faces and license plates are pixelated; people may still be visible, for example by their clothing. Images that are not anonymized are never stored, transferred or published. The anonymization runs on a computer in the vehicle that is operated by KE:SAI itself.

Automatic detection is highly accurate but cannot be guaranteed to be complete. If you find a recognizable face or license plate in our data, please inform us at info@kesai.eu, and we will correct or delete the affected images.

We use the recorded data for research on and development of self-driving systems. In particular, we use it to train and evaluate driving policies and generative models of driving scenes. Since the models are trained only on anonymized images, they do not learn the faces and license plates that were pixelated. In addition, we publish parts of the anonymized data as open datasets that anyone may use, including companies for commercial purposes. Within a recording, our software detects road users and tracks them over consecutive frames, for example to estimate their speed. This tracking uses only the anonymized images. We do not identify any person, do not recognize people across different recordings, do not create profiles and do not link the data to other information about individuals.

The recording and anonymization of the data, its storage and use for our research, and the publication of anonymized datasets are based on our legitimate interest pursuant to Art. 6(1)(f) GDPR. Our legitimate interest is research on safe self-driving systems and making open driving data available to research and industry. Your interests do not override this interest, because faces and license plates are anonymized before any data is stored, only anonymized images are ever published, recordings are removed on request and excluded from all future releases, and we neither intend nor attempt to identify any person.

Recordings on closed test sites, such as driver training grounds, are not sufficient for this research. Self-driving systems must handle the variety of real traffic: many different road users and their interactions, road layouts, traffic densities, weather and lighting conditions, and rare events. Closed test sites cannot reproduce this variety, and systems developed only there do not transfer reliably to public roads.

The anonymized data is stored on servers in Germany and France. As part of our open research, we publish anonymized data as open datasets and use it in public research challenges. Published datasets are available to anyone worldwide, and their license allows any use, including commercial use. We may also share anonymized data with research partners. Data that is not anonymized is never shared.

We keep the anonymized data for as long as we need it for our research. At least every 3 years, we review whether it is still needed and delete data that is no longer needed. Published datasets remain publicly available without a time limit. Images that have not been anonymized are not stored.

We do not carry out automated individual decision-making or profiling within the meaning of Art. 22 GDPR.

Under the GDPR, you have the right of access (Art. 15), rectification (Art. 16), erasure (Art. 17), restriction of processing (Art. 18) and to object to the processing (Art. 21). Since we do not identify the people in our data (Art. 11 GDPR), we need your help to find a recording: please tell us the place, the date and the approximate time at which you were near one of our vehicles, and if possible a description that helps us find you in the images. Send your request to info@kesai.eu.

On request, we delete the affected recordings from our own data and exclude them from all future releases of our datasets. We cannot delete copies that others have already downloaded. When we remove recordings on request, we note this in the release notes of the next dataset version.

You also have the right to lodge a complaint with a data protection supervisory authority, in particular in the country of your habitual residence, your place of work or the place of the recording. The authority responsible for KE:SAI is the State Commissioner for Data Protection and Freedom of Information Baden-Württemberg (Lautenschlagerstraße 20, 70173 Stuttgart, Germany, baden-wuerttemberg.datenschutz.de). In France, you can contact the Commission nationale de l'informatique et des libertés (CNIL, cnil.fr).

Status: November 2026