In recent years, translation quality evaluation has emerged as a major, and at times contentious, topic. The industry view on quality is highly fragmented, in part because different kinds of translation projects require very different evaluation methods. In response, the EU-funded QTLaunchPad project has developed the Multidimensional Quality Metrics (MQM) framework, an open and extensible system for declaring and describing translation quality metrics using a shared vocabulary of “issue types.” As an effective approach to evaluating AI translation quality, the classification of translation errors has drawn increasing attention. This study focuses on translation errors in red texts, using the MQM quality assessment model as the analytical framework to categorize errors in translations produced by ChatGPT4.0, a leading engine among current large language models. The findings aim to provide pedagogical support for pre-editing and post-editing training in professional translator education.