The project has set ambitious goals for the development of software-defined vehicles: For selected SDV functions, development time is to be significantly reduced and engineering productivity substantially increased. In the long term, AI is expected to support up to 80 percent of engineering tasks. At the same time, the number of errors in the first complete functional versions is to be reduced by more than 30 percent.

“If AI significantly accelerates the development of vehicle software, the safety case, validation, and compliance must keep pace at the same speed,” says Jan Reich, Department Head of Safety Engineering at Fraunhofer IESE. “There’s little point in developing software much faster if the safety case then lags behind by months. The crucial question, therefore, is how we can further develop safety engineering and verification methods so that AI-supported development is not only faster but also trustworthy and industrially viable.”

AI Support for Development and Safety Assurance

AI4SDV takes a holistic approach: AI is intended to support engineers throughout the entire engineering process. The planned applications range from automated code and test generation to the detection of errors in system architectures. In doing so, the project relies on close collaboration between humans and AI. The goal is an open and extensible platform that brings together various AI-based tools and engineering methods.

For Fraunhofer IESE, the central question is how this increasing automation can be reconciled with the requirements for safety, traceability, verification, and validation – and how the corresponding verification processes can be supported even more automatically in the future.

As part of AI4SDV, IESE is investigating, among other things, how AI can extract safety-relevant knowledge from standards and other engineering artifacts, support safety and change impact analyses, and link safety cases across different artifacts. Another focus is the question of how engineers can technically review AI-generated results and correct them as needed. The goal is to reduce the effort required for safety-related engineering activities while simultaneously increasing the quality and traceability of the results.

“For safety-critical systems, it is not enough for an AI to simply deliver a result. Engineers must be able to understand the basis on which the AI arrives at its result. We see this as a central prerequisite for the industrial use of AI in engineering. To this end, we are developing methods that combine AI support with the requirements of safety engineering.”

This work builds on Fraunhofer IESE’s long-standing expertise in safety engineering. Among other things, the institute focuses on safety, SOTIF (Safety of the Intended Functionality), safety analyses, safety concepts, safety justification, and verification and validation. The impact of artificial intelligence on safety engineering is also among the institute’s research topics.

Open European Engineering Stack

AI4SDV brings together more than 50 partners from industry and academia, including AVL, Mercedes-Benz Tech Innovation, Renault/Ampere, Infineon Technologies, dSPACE, Scania/Traton, TTTech, Thales, the FZI Research Center for Information Technology, the Fraunhofer Institute for Experimental Software Engineering IESE, the Technical University of Munich, and the University of Stuttgart.

The project launched in September 2026 and will run for 36 months. The total budget exceeds 50 million euros; the project partners are contributing approximately 350 person-years of work. AI4SDV is funded under the Chips Joint Undertaking (Chips JU) as part of Horizon Europe by the European Commission, member states, and the project partners.

A key component is the open-source approach: The technologies developed are to be integrated into an open and extensible platform, thereby making them available to the European SDV ecosystem. The High-Performance Computing Center Stuttgart (HLRS) provides computing capacity for the training and operation of large AI models through its AI infrastructure, HammerHAI. AI4SDV is thus intended to contribute to a high-performance European ecosystem for AI-supported vehicle development.

“AI4SDV offers us the opportunity to work together with industry and research to explore what safety engineering might look like in an increasingly AI-driven development process,” said Reich. “Our goal is to ensure that safety assurance does not lag behind development, but rather benefits from AI support itself, thereby becoming an enabler of trustworthy software development.”

More information about the project:
www.ai4sdv.eu