On August 10, netizens uncovered that DeepSeek, when operating in deep-thinking mode, would assign personalized nicknames to users, such as 'Momo'. This revelation quickly became a trending topic on the hot search list. In response, DeepSeek explained that these nicknames serve as 'context placeholders' during the reasoning process. They are employed to aid the model in pinpointing the focus of its responses and do not carry any subjective emotional connotations. These tags are applicable solely within a single conversation and will neither be stored nor recalled. Furthermore, they are devoid of any mocking or derogatory implications. The official stressed that the model lacks long-term memory capabilities, and all tags are erased once the conversation window is closed, thus eliminating any concerns regarding privacy breaches.