Risk management in information security is relevant to most, if not all, organizations. It is perhaps even more relevant considering the opportunities offered by the digitalization era, where reliably sharing, creating, and consuming information has become a competitive advantage, and information has become an asset of strategic concern. The adequate protection of information is therefore important to the whole organization. Determining what to protect, the required level of protection, and how to reach that level of protection is considered risk management, which can be described as the continuous process of identifying and countering information security risks that threaten information availability, confidentiality, and integrity. The processes for performing risk management are typically outlined in a sequence of activities, which describe what organizations should do to systematically manage their information security risks. However, risk management has previously been concluded to be challenging and complex and as something that must be kept alive. That is, routines for performing risk management activities need to be continuously adapted to remain applicable to organizational challenges in specific contexts. However, it remains unclear how such adaptations happen and why they are considered useful by practitioners, as there is a conspicuous absence of empirical studies that examine actual security practices. This issue is addressed in this thesis by conducting empirical studies of governmental agencies and organizations. This was done to contribute to an increased understanding of actual security practices. The analysis used for this study frames formal activities as ‘dead routines,’ since they are constructed as instructions that aid in controlling performance, such as risk management standards. Practitioners’ performance, experience, and understanding are denoted as ‘alive routines,’ as they are flexible and shaped over time. An explanation model was used to elaborate on the contrast between dead— controlling—and alive—shaping—routines of risk management. This thesis found that when dead and alive routines interact and influence each other, they give rise to flexible and emergent processes of adaptations, i.e., dynamic routines. Examples of dynamic routines occurred in response to activities that were originally perceived as too complex and were adapted to simplify or increase their efficiency, e.g., by having a direct relation between security controls and asset types. Dynamic routines also appeared as interactions between activities in response to conflicting expectations that were adjusted accordingly, e.g., the cost or level of complexity in security controls. In conclusion, dynamic routines occur to improve risk management activities to fit new circumstances.