Meeting safety objectives of Advanced Driver Assistance Systems (ADAS) and Autonomous Vehicles (AVs) in all driving situations remains a key engineering challenge. The tight integration of system design, safety analysis, software/hardware design, simulations of driving scenarios for early validation and verification activities into a fully digitalized workflow is an enabler to compile the overall safety case incrementally and in a consistent manner. We present an agile and model-based approach that combines system and safety engineering in a digital environment. We use the SysML V2 language throughout the lifecycle to define system models, on which we build the safety analyses and the validation through simulation. This is repeated iteratively on each abstraction level, enabling early feedback and evolutionary improvement. To efficiently create a consistent safety argumentation that can keep pace with the frequent extensions and adjustments, we exploit the model-based engineering results and the relations between them as evidences in a dynamically generated safety case based on the Goal Structuring Notation (GSN). This way we can also support change impact analysis based on the propagation of changes along the relations in the model. Conceptual proposals for safety management and the coupling with safety engineering are presented and the process is demonstrated by the example of a highway-pilot system.