Overfitting

When a model learns the training data too well, including its noise and quirks, and performs poorly on new, unseen data. An overfitted model has essentially memorized the training examples rather than learning the underlying patterns, making it brittle in real-world use.
See also: generalization, regularization, early stopping.