Imagine training a powerful AI model without the tedious manual labor of labeling data. That's the exciting promise of a new 'teacher-student' framework, where a large language model (LLM) acts as the teacher, automatically classifying news articles by topic. Researchers explored this concept using the IPTC Media Topic schema, a standard used by news providers worldwide. They used a powerful GPT model to annotate a massive dataset of news articles in Slovenian, Croatian, Greek, and Catalan, effectively teaching a smaller, more efficient XLM-RoBERTa model (the student). The results? The student achieved accuracy comparable to its teacher, even with far less training data, demonstrating the effectiveness of this automated approach. Even more impressive, the student model displayed remarkable 'zero-shot' cross-lingual abilities, accurately classifying news in languages it hadn't explicitly seen before. This research opens doors for building highly efficient, multilingual AI models that could revolutionize news classification and beyond, eliminating the bottleneck of manual data annotation and paving the way for more scalable AI solutions in the future. While challenges remain in disambiguating overlapping topic categories, this approach is a significant step towards automating AI training and unlocking the potential of smaller, more accessible AI models.