BMW Group advances use of AI in crash simulations.

Today, vehicle safety begins in the digital realm. The BMW Group and Mistral AI are advancing the use of artificial intelligence (AI) in crash simulations. The aim is to analyse virtual crash tests faster, hone development work and enhance passive vehicle safety through targeted support.

Physical crash tests build on extensive virtual simulations. Before a new car ever hits a barrier in real life, it has already undergone myriad impact scenarios in the digital realm. Each week, the BMW Group carries out thousands of virtual crash simulations, generating insights that allow vehicle structures to be evaluated much earlier in the development process.

This produces vast amounts of data capturing material behaviour, deformation patterns and the distribution of impact forces throughout the vehicle. Over time, a dataset of historical proportions has evolved: a memory built from countless accident simulations that now exceeds one petabyte.

It is this wealth of data that the BMW Group and Mistral AI now aim to leverage even more fully. The two companies are partnering to further advance the use of artificial intelligence (AI) in crash simulations. Their goal is to harness AI to analyse vast volumes of simulation data more rapidly and identify patterns at an earlier stage and enable more precise, data-driven development decisions.

Dr Franz Decker, CIO and Senior Vice President with the BMW Group, explains how this works in practice: “By combining our engineering datasets with Mistral AI’s model training capabilities, we are building a specialised AI that supports complex development tasks.”

Crucially, the BMW Group aims to use AI models that have been specially trained for industrial applications and therefore possess capabilities that extend far beyond those of regular AIs. Known as Large Industry Models (LIMs), these are trained on industry-specific engineering and simulation data. Unlike generic AI systems, LIMs have domain expertise embedded directly into them to provide exactly the technical expertise vehicle development and crash tests depend on.

This is particularly relevant for crash simulations, as these involve a highly complex interplay between physics, material behaviour and data analysis. The specialised LIM can identify where structures absorb energy, how materials behave under stress and which anomalies warrant closer examination, for example.

With AI as their copilot – but not their replacement! – engineers will be able to navigate complex data landscapes faster and in greater depth.

Mistral AI, too, sees the project as a powerful example of industrial AI: “As industrial AI becomes the new frontier, we are proud to partner with the BMW Group,” says Marjorie Janiewicz, Chief Revenue Officer at Mistral AI. “This collaboration shows how industry-specific AI models can help solve complex engineering challenges such as crash simulation.”

For the BMW Group, the collaboration with Mistral AI also marks a starting point for deploying application-specific AI solutions in other areas of vehicle development and along the value chain. Virtual crash tests serve as a key lever for the future of development – accelerating processes, enabling precise, analysis-based decision-making, and laying the foundation for high-quality, data-driven value creation.