In her February 2020 column ("Are You Sure Your Software Will Not Kill Anyone?"), Nancy Leveson says the solution to software safety is not "building a software architecture and generating the requirements later." Reading this, we were surprised that anyone would propose such an approach, or that Leveson would find it necessary to argue against it. We were even more surprised to see that Leveson attributes the proposal to a National Academies report that we edited.1
Perhaps Leveson had a different report in mind. Our report actually substantiates the very arguments she makes in her column, identifying the same misconceptions that she notes, along with others—and provides citations for the earlier origins of these supposedly new arguments.