Autonomous Driving: Valeo to Release AI Code and Data as Open Source
Valeo has decided to fully publish the code, trained models and datasets of a future autonomous driving system, developed with a non-profit research laboratory. Announced on October 13, 2026 in Paris, this Franco-German initiative aims to provide automakers, suppliers and researchers with a common reference to evaluate their own systems.
A Research Collaboration Around Generalist Autonomous Driving AI
Valeo and KE:SAI (Kyutai ELLIS Scalable Autonomous Intelligence) have announced a research collaboration in open science focusing on the development of a driving system for autonomous vehicles based on generalist and fully reproducible artificial intelligence.
KE:SAI is a non-profit open science research laboratory, co-founded by Kyutai and the ELLIS Institute Tübingen. On Valeo's side, the project is led by Valeo Brain, whose Chief Executive Officer, Marc Vrecko, commented in the press release.
The program sets a two-stage objective: to equal, then improve, the performance measured on the main public benchmarks for autonomous driving evaluation.
Code, Trained Models and Data Published in Open Access
All project components (code, trained models and datasets) will be published as open source. Results will also be shared through publications.
This choice will allow automotive manufacturers, suppliers and researchers to reproduce results, build upon the model and compare their own systems to a common reference. The two partners present this approach as a foundation for democratizing autonomous driving technology and physical AI.
An Open Approach Applied to Physical AI
Andreas Geiger, CEO of KE:SAI, justifies the chosen model by drawing a parallel with language models: "For language models, open-source enabled years of progress in just a few months. Yet cutting-edge physical AI remains largely closed." He adds: "Everything we develop will be open."
For Marc Vrecko, the collaboration "marks a strategic milestone for Valeo, and for the European automotive ecosystem as a whole".
The announcement comes shortly after Valeo's selection by Nissan, made public on October 1, 2026, for the next generation of its ProPilot assisted driving system. This agreement integrates the autonomous driving AI technology developed by Wayve within an automotive quality system.
