AI, as it exists today, is very good at interpolating lots of data. It is far less able to do a good job of extrapolating far outside the boundary of existing d

AI, as it exists today, is very good at interpolating lots of data. It is far less able to do a good job of extrapolating far outside the boundary of existing data. Given the cat-and-mouse nature of security, where the parties regularly move outside the bounds of past behaviors, it is very very unlikely that AI will be useful in security at the boundary. It will be very good at making sure known exploits are not available, but known unknowns and unknown unknowns will remain security blind spots.