A mathematical framework for guaranteeing that the output of an algorithm reveals little about any individual data point in the training set.
Differential Privacy (DP) adds calibrated statistical noise during the training or querying of AI models, so that including or excluding any single record has only a small, bounded effect on the output. For Private AI deployments that handle PII or PHI, DP is a safeguard against model inversion and membership inference attacks.
See also Private AI, Air-Gapped Environment, Role-Based Access Control and Zero Trust Architecture.