Maching learning and artificial intelligence has been in the making since Alan Turning.  Only recently has it gone mainstream and it feels that everyone is getting on the act.  Including the usual snake oil sellers who make claims that the underlying data that these algorithms rely on.  Developers have to appreciate there is a societal element that does not take kindly to biased or discriminatory output.  In short they need model risk management, a new frontier which will not come online until someone makes a catastrophic mistake.