Chinese large language models exhibit privacy protection shortcomings, necessitating further optimization to minimize privacy risks in applications.