What are the drawbacks of reusing model classes for API testing?

does it make sense to use existing app class models (or DTOs) for API testing? Or should the testing code use its own models? I'm thinking that reusing app models might hide some bugs, but it's hard to estimate the risks.