Distribution shift — Distribution shift happens when the data a model sees in real use differs from the data it was trained on. A model trained on summer photos can fail on winter scenes. It is a major reason AI performance drops after deployment. Monitoring and retraining help address it.

Added October 21, 2026