There are no such models, depending on your definition of censorship. If you're referring to abliteration and similar automated techniques, they're snake oil.