Bayesian networks are white-box AI models that encode uncertain causal relations using stochastic variables and include intuitive graphical structures and associated conditional probability tables. Such a network typically includes administrative and medical patient-specific information (e.g., age, sex, smoking history); symptoms (e.g., biomarkers, CT scan results); decision making variables (various treatment options), and outcome variables (e.g., disease-specific survival rate as well as side effects). These networks are typically developed using a combination of expert knowledge and data. The main problem addressed in this work package is how accountability can be facilitated by traceable development and annotated inference and explanation in (variants of) Bayesian networks. We thus focus on the technological advances necessary to realise accountability in the usage of a Clinical Decision Support System.