Open Models Almanac
Methods

single mention · not yet filed

Input/output moderation classifiers layered around the model (defense-in-depth)

The sources on this page only mention it in passing.