This is cool because it kind of gets at the core of transparency about AI models.
I would imagine if this would built a certain way it could not only tell a brand what I models are saying, but could be used to look for bias in models that do credit scores, for example.
P.S. I also used to work at Rover!