A study reveals that the DeepSeek AI chatbot censors 85% of the prompts related to 'sensitive topics' for China.