Recent rapid advances in ML have made the potential power of AI blindingly obvious. What’s much less obvious is how it is going to be usefully deployed in security contexts and whether it will deliver the major breakthroughs its proponents promise.