In many aspects, human actors and their interactions with technical systems differ from the interactions between technical systems and components. Consequently, human actors should be treated differently in early stages of system design. This is even more important when the system of interest contains subsystems with artificial intelligence. SysML-based approaches to model and assess special aspects of human-system interaction will be presented. It will be shown how this information can be used for early stages of risk assessment and risk management. It is suggested that components containing artificial intelligence can be treated with similar approaches and modeled in ways which clearly differentiate them from conventional technical components.