I can see no problem here if it is objectively proven to be capable of reliably confirming whether a candidate is an expert. Unless you artificialy rule that ca

I can see no problem here if it is objectively proven to be capable of reliably confirming whether a candidate is an expert. Unless you artificialy rule that candidates can only apply once. For cases when a candidate emerges who believes he somehow really can't pass the AI-powered test despite being perfectly competent there obviously should be fallback alternatives available, i.e. a jury of experts who would talk to him and tell whether there is a chance his claim is legitimate or he's just kidding.