The article makes the case that if an interoperable model (one we can explain) isn't used, then the user of the black box model should have the burden of proof to prove that no interoperable model exists that does the job, and some level of responsibility for trying to develop one.