This paper provides an empirical and conceptual account on seeing machine learning models as part of a sociotechnical system to identify relevant vulnerabilities emerging in the context of use. As ML is increasingly adopted in socially sensitive and safety-critical domains, many ML applications e…