So, DeepSeek is actively censoring controversial prompts or is this just a fun execise?