Automated driving has evolved over nearly a century, from early infrastructure-guided prototypes to today’s vehicle-centric, AI-enabled systems. Each stage in this development has been driven by the pursuit of safe, clean, efficient and accessible mobility yet has also raised fundamental questions about how the safety of automated vehicles (AVs) can be assured, demonstrated and trusted.

This paper, produced by the FISITA Automated Driving Expert Group, traces the historical progression of
automated driving, outlines the current regulatory and standards landscape and explores how data can
underpin credible, evidence-based safety assurance. As automated driving technologies transition from
rule-based architectures to complex, end-to-end deep learning systems, conventional approaches to proving safety through accumulated mileage or human-driver comparisons are no longer sufficient. Instead, the assurance process must shift toward continuous, data-driven evaluation.
AVs generate multiple categories of data that are essential for this process:

• Operational data captures vehicle dynamics, system status and environmental conditions.
• Perception and decision-making data records how the vehicle senses, interprets and responds to its
  surroundings.
• Event data provides detailed context for incidents, disengagements and near misses.
• Training data supports the development and refinement of machine learning algorithms.

Together, these datasets offer a rich source of insight into system performance but their regulatory use is
complicated by the diversity of hardware, software and data architectures among developers. The absence of standardised formats and common interpretive frameworks limits transparency, comparability and effective oversight.

To address these challenges, the British Standards Institution’s (BSI) Digital Commentary Driving (DCD)
framework proposes a standardised method for capturing and communicating how AVs perceive and decide, inspired by the human practice of commentary driving. By defining a consistent structure for explaining system reasoning, DCD provides data that enables examination of AV perceptions, predictions and behaviours to assess safety performance without exposing proprietary code, thereby enabling more transparent regulatory evaluation and public accountability.

This paper concludes that data transparency is both a technical and societal imperative. Building public trust in automated driving will depend on establishing shared frameworks for data collection, interpretation and communication, ensuring that automated vehicles not only operate safely but can be seen to do so. At the same time, consumers may need to accept confidential collection and sharing of data of their use of AVs to support continuous improvement of the safety and performance of such services. Through responsible data use and collaborative safety assurance, the vision of safe, clean, efficient and accessible mobility can move from aspiration to reality.