Could this be used to infer the alignments done by the creators of the models by passing in a common set of questions to before and after and then comparing the

Could this be used to infer the alignments done by the creators of the models by passing in a common set of questions to before and after and then comparing the results? Would be interesting to see what Elon has done to his XAI model in comparison to OpenAI.