The systemn I work with, AMBER, explains how individual classical terms are derived: https://ambermd.org/tutorials/advanced/tutorial1/section1.ht... which appears te be MP2/6-31g* (sorry, I never delved deeply into the QM parts). Once those terms are derived, along with various approximated charges (classical fields usually just treat any charge as point-centered on the nucleus, which isn't great for stuff like polarizable bonds), everything is purely classical springs and dihedrals and interatomic potentials based on distance.

I am more than happy to use "ab initio" purely for QM, but unfortunately the term is used widely in protein folding and structure prediction. I've talked exdtensively with David Baker and John Moulton to get them to stop, but they won't.