A condition whereby an AI model is not generalized sufficiently for all uses. Although it does well on the training data, overfitting causes the model to perform poorly on new data.
Overfitting can result when there are too many epochs (training phases), too little data examples or the model is simply too complex to begin with. Overfit models are said to "memorize" rather than "generalize" and be like a product that does well in a lab but not in the real world. See
AI programming.