This thing is not much more than a database of all the data it has seen. Far from "just like a human who reads various books". If they want to make it useful for "humanity" they should release it under GPL - probably that would be complicated given all the mix of the licenses used though.

This is in essence similar to Clearview AI. Data/code is published with an specific intention, which is reflected by the license in case of code. It should not be used outside of that intention which for GPL is clearly stating that the author does not want to have their code used as part of a commercial software.