It's incredibly concerning that we could have "medical students and other health professionals" reading images and providing labeling data that would actually get used to teach AI...
The way we handle this is that we only trust our users based on their past performance on "gold standard" cases, regardless of their credentials. So if a med student (or even a layperson!) looks at cases on the app, they aren't contributing to labeling data that we provide to customers until we've learned to trust them based on their performance on hundreds or thousands of past cases. At the same time, we don't trust people just because they're a doctor if they don't perform well at the task :)