I tried to build something: https://github.com/YM2132/PR_guard which aims to help in these cases. It's not perfect but with stronger AI detection tools (Pangram) it could be improved although the issue of cost then arises and who pays for it.