BMW says data collection is triggered only by defined driving situations. Possible triggers include interventions by assistance systems, heavy braking and sudden evasive manoeuvres.BMW / Uwe Fischer
Real-world driving data is becoming a key input for developing and validating increasingly complex driver-assistance functions. BMW is expanding event-triggered image and sensor collection from consenting customer vehicles in Europe to feed machine-learning development.
BMW is expanding the collection of image and sensor data from customer vehicles in Europe. The carmaker plans to use the material to improve driver-assistance systems and partially automated driving functions, provided individual users have consented to the data collection.
Since April 2026, BMW has been able to capture images from real traffic situations in vehicles from its latest model generation. Eligible cars need the required sensor suite and a suitable data-processing architecture. The planned roll-out includes BMW’s Neue Klasse iX3, the i3, X5 and 7 Series, with further new and updated models expected to follow. The customer-fleet data adds another layer to BMW’s existing development methods. Assistance functions are already tested through virtual simulation and development vehicles, while regular road use can expose situations that structured test programmes may not reproduce as frequently.
BMW says the system is not designed to record continuously. Instead, data collection is activated by defined events in real traffic. These can include an intervention by an emergency-braking system, heavy manual braking or a sudden evasive manoeuvre. A recording can also be triggered when a driver-assistance function prevents a potential collision during a lane change.
Advertisement
According to BMW, the vehicle captures only information intended for analysing the relevant situation. Images from exterior cameras can be combined with data from other environment sensors as well as vehicle-dynamics parameters such as speed, direction of travel and steering angle.
From mid-September 2026, event-triggered video data can also be transferred. BMW says each recording is limited to a maximum of 120 seconds. The video initially reaches the company’s IT systems in its original form. The approach illustrates how connected fleet data can support assisted-driving systems by adding real-world edge cases to information generated in controlled development environments.
BMW intends to feed the collected information into machine-learning processes used to develop driver-assistance systems and partially automated functions. Improvements derived from this work could later be delivered to suitable cars through software updates.
That gives the production fleet a more active role in the continuing software-development cycle. Real driving situations can show how assistance systems behave across different roads, traffic patterns and environmental conditions. The event-based data therefore creates an additional source of engineering evidence alongside simulation and conventional vehicle testing, with the value of the dataset depending on how well it captures relevant and sufficiently varied scenarios.
How does BMW address data privacy?
Customer consent is a prerequisite for the collection, according to BMW. The company also points to a privacy-by-design approach. It says the vehicle identification number is deleted immediately after the data is transferred to the backend, preventing a recording from subsequently being assigned to a specific vehicle.
BMW says its systems are not intended to identify individual road users. Where employees need to access particular recordings for development work, recognisable faces and number plates of other road users are to be obscured, as far as technically possible, before the footage is displayed. The measures reflect broader privacy questions around camera-based driver assistance as vehicles generate and process increasing volumes of sensor data.
Advertisement
By expanding event-based collection, BMW is tying the evolution of its assistance software more closely to evidence from customer vehicles in everyday traffic. How strongly that information influences individual functions will ultimately depend on the volume and quality of the recordings and on whether they capture situations that meaningfully extend the company’s existing validation data.
BMW CIO Decker at automotiveIT Kongress 2026
The growing use of vehicle data shows how closely software development, AI and IT infrastructure are now linked to the evolution of the vehicle itself. Dr Franz Decker, CIO and Senior Vice President Group IT at the BMW Group, will explain the role that global platforms and artificial intelligence play within the company at the automotiveIT Kongress 2026 in Ingolstadt.
In his presentation on 14 October, Decker will discuss the operating model of BMW Group IT. The focus will be on scalable platforms, AI agents for automating business processes, as well as cybersecurity and resilience. According to BMW, these technologies are intended to create measurable value across the automotive value chain.
