Calling ML as AI itself is hyperbole. ML is just a set of algorithms (or extensions of statistical modeling) each of which act like kinds of expert systems (only good to solve certain problems efficiently), usually many of them are implemented in Python due to plenty of useful libraries. But we still know almost nothing about how human think. Not all ML require training set but I don't know what algorithms are used in security industry. The reason ML is massively used in security is probably because processing such a big data (either thousands of malware or massive logs) by human costs too much or impossible, so it's matter of efficiency and not accuracy. You may also want to question what data will be collected and how your privacy is protected. We know why Google invest so much into ML and big data, and those data might be abused by 1 evil employee if not properly protected (But even w/out such abuse, is it really no problem at all?).
by Nicholas Carlini , Chang Liu and Dawn Song each of University of California, Berkeley.....Jernej Kos of National University of Singapore.....Ulfar Erlingsson of Google Brain
Enlightening. What have been learned in NN is not clear so a user might forget about them, but they are there.