The Role of CE Marks in AI Interventions

The CE mark is basically clearance to be put on the market in Europe. So we don't change the model itself. We may do validation to make sure that it works in the population. And we also do updates, right? If we understand how to make the model better. But those are very thought out and we file requests to change it with notified bodies before we actually change it. Are you allowed to reuse the often laboriously collected validation data that was used in validating the regional model? I think this is a quickly, people are thinking about this in terms of how many times is it just once? Is it twice? When are you going to start overfitting?