A machine learning task where a model reads text and assigns it to predefined categories, such as 'safe' or 'unsafe'.