As a methodology for system design and testing, use cases are well-known and widely used. While current active machine learning (ML) algorithms can effectively automate unit testing, they do not scale up to use case testing of complex systems in an efficient way. We present a new parallel distributed processing (PDP) architecture for a constrained active machine learning (CAML) approach to use case testing. To exploit CAML we introduce a use case modeling language with: (i) compile-time constraints on query generation, and (ii) run-time constraints using dynamic constraint checking. We evaluate this approach by applying a prototype implementation of CAML to use case testing of simulated multi-vehicle autonomous driving scenarios.