The nature of AI is a tradeoff between (arguable) results and/or convenience vs. losing the how of the solution (by design) and losing the human training value of working out the solution.
Unless I'm mistaken, both of these losses are intrinsic to AI. Also, they are precisely the two things that these math challenges were designed to produce.
So it would seem that the decision-making should be straightforward. Require anyone claiming a prize to show every step in the solution, whether they use AI or not. No black box answers.