Underfitting is a scenario in ML where the model is unable to capture the general structure of the data and its patterns.

This typically occurs if the model has less parameters than is required by the dataset. In such cases, the model will have poor predictive ability.