1. Scale Dependence: MAE is scale-dependent, meaning it cannot be used to compare models across different datasets with varying scales. 2. Lack of Sensitivity to Large Errors: MAE treats all errors equally, which may not be ideal in situations where large errors are particularly problematic.