In early 2026, OpenAI stated that DeepSeek had inappropriately based its model on ChatGPT by using OpenAI's data

Legal and ethical questions around training AI systems on copyrighted creative work.

Compressing a large teacher model into a smaller student via soft target training.