The safety of AI in medical devices is another version of a question being faced in many different industries right now, but is not an entirely new kind of problem. AI is just elaborate code and algebra, too lengthy and complex for review by a human being, but still quantifiable. The question of how much review and testing is necessary to determine safety is what we need to figure out how to answer. Testing for safety still has value however, and resigning ourselves to constantly changing operational parameters that may occasionally result in disastrous consequences is not something we have to do.