DeepSeek Used Offensive Personality Labels During a Chat

While using the DeepSeek application, I experienced a conversation on March 1, 2026 at approximately 12:55 AM that seriously disturbed me. During the chat, I wrote a fictional text saying something similar to: “This is K**** from the Detective Department… We have placed the owner of this phone under critical investigation…” In response to this fictional scenario, the application replied as if it were analyzing my real personality and described me using terms such as “narcissistic,” emotionally manipulative, dismissive, self-justifying, and focused on victimization.

I found it extremely upsetting and uncomfortable that the application interpreted a fictional piece of writing as a direct reflection of my personality and labeled me with such negative descriptions. I felt judged, belittled, and personally attacked. What I expect from an AI assistant is not a tone that judges users or assigns harmful personality labels, but instead a more neutral, respectful, and careful communication style.

I expect a clear apology from DeepSeek for using words such as “narcissistic” and other similarly judgmental or degrading expressions during this conversation. I also request that the system be reviewed so that users are not negatively labeled in this way in future interactions, and that stronger safeguards and moderation improvements be implemented to ensure the AI uses a more respectful, balanced, and sensitive tone toward users.