Why wouldn’t we know? This is a statement with no explanation or justification. No one should be expecting it to be taken as an axiom or a given. For that reason I don’t accept it as a legitimate argument nor would most people. There is no basis in reality to believe that we would not review code written by any AI. In fact currently most code written by AI is inherently broken and incomplete requiring human intervention to make it even function. There is also no evidence of any intent, nefarious or otherwise on the part of current AI systems. AI systems goal seek they don’t have intent. Could it happen someday? Perhaps but that’s pure speculation. Making speculative and fantastical statements about technology of the future that is far beyond current capabilities isn’t a discussion about the current or near term challenges facing society adapting to these technologies, it’s science fiction.