Liver failure is a severe liver injury caused by multiple factors， leading to significant impairment or decompensation of the liver’s synthetic， detoxification， metabolic， and biotransformation functions， and it is a group of clinical syndromes with the main manifestations of jaundice， coagulation disorder， hepatorenal syndrome， hepatic encephalopathy， and ascites. Early and accurate prognostic prediction is crucial for improving the clinical outcome of patients. In recent years， artificial intelligence （AI）-based early warning and prediction models are gradually transforming the traditional diagnostic and therapeutic approaches. This article systematically reviews the advances in the application of machine learning-based early warning and prediction models in acute liver failure and acute-on-chronic liver failure， and related models have shown good performance in risk stratification and prognosis prediction. With the continuous development of related technologies， AI is expected to provide new opportunities for the early intervention and precise treatment of liver failure， thereby significantly improving the prognosis of patients.