What did they train the AI on to deem something 'extremist'?

Should we get to see the training data used and labels?

Or is this the modern day equivalent of credit score algo, something that can have huge impact on lives, but you are not allowed to know what it is.

This is bad.