I think the reasoning is innocent, the problem is that if it gets leaked you have people (probably incorrectly) interpreting things using the data which is half the problem.
"Your ethnicity preference is white? You goddamn racist".
This gets worse when those interpreters are in a position of power.
