The purpose of this paper is to investigate how to best learn to run a collaborative robot in a safe way through different types of vocational training. To learn how to run a collaborative robot safely has many challenges. The challenges connected to learning are: i) life-long learning and re-skilling takes time to deploy and ii) in traditional training information is forgotten. In addition, there is a lack in labour and although Industry 4.0 envisioned an operator that would be intelligent and skilled to manage complicated systems there simply is no such personnel. Therefore, vocational training should be designed to fit everyone. Three experiments were carried out: 1) how safe the operator felt and what type of theory that was preferred, 2) if and how VR could be used for safety-critical tasks and 3) how a reduced version of theory and how having a training online effected the operators experience of feeling safe. Results indicate that theory length must be adapted to the trained task, that operators feel safe due to the presence of the trainer (both IRL and virtual) and that VR could be used to learn safety-critical aspects in an efficient way. In conclusion, this article shows that safety-critical tasks could be trained through VR and that the design of the vocational training in this article is a way to ensure operator safety and the perception of being safe when working with a collaborative robot.