The safety validation has proven to be one of the most obstructive challenges in the pursuit of highly automated driving. Exhaustive field tests have been deemed infeasible and novel approaches such as scenario-based testing yet require to be proven viable. The challenge becomes even more serious when considering changing architectures due to learning software components and over-the-air updates. A modular approach to safety approval, focusing on assuring the safe operation of individual modules in their respective environments, promises to reduce the overall effort. In particular, it provides an argument for preserving the approval for future updates and upgrades, reducing the need for intensive retesting of the whole system. However, established knowledge-based methods for decomposition, specification, risk analysis and module test case generation struggle to argue completeness. Data-based methods used in other applications provide the opportunity to support this issue. Inspired by the design-by-contract paradigm, we combine selected methods into a framework to feed a data-driven interface definition to support modular safety approval. The framework is practically applied in a simulation environment to a highly automated vehicle with a disruptive modular architecture. A reduced ODD setting serves as a proof of concept and provides insights on the limitations and applicability of the applied methods with respect to the use cases of a modular safety approval.