Unfortunately, the nature of the models is that it's very difficult to get an understanding of the innards. That's part of the point, you don't need too. The best we can do is monitor how it's built and what connects in and out of it.

The open source bits let you see that it's not passing data on without your permission. If the training data is also open source, you can get for biases e.g. 90% of faces being white males.

No amount of ML expertise will let someone know how a model produced a result, exactly. Training the model from the data requires a lot of very delicate math being done uncountable times to get a model that results in something useful, and it simply isn't possible to comprehend how the work inside is done in a meaningful way other than by doing guesswork.

We already know how these models fundamentally work. Why exactly does it matter how a model produced some result? /gen