Abstract: Modern safety-critical organizations are characterized by complex, nonlinear dynamics involving many interrelated actors and processes. Safety issues that emerge from these complex dynamics increasingly remain hidden, until an incident or even a serious accident occurs. Traditional safety analysis methods developed long ago for much simpler organizations cannot help identifying, explaining and predicting many safety-related properties of modern organizations. To address this issue, in the paper a formal approach is proposed to establish relations between local dynamics of actors of a complex safety-critical organization and global safetyrelated properties that emerge from these dynamics. In contrast to the traditional approaches, the organizational dynamics are specified by taking the agent perspective with an organizational layer. The application of the approach is illustrated by a simulation case study, in which spread of safety-critical information in an air navigation service provider is investigated. © 2012 ISCRAM.