-
Notifications
You must be signed in to change notification settings - Fork 1k
Open
Description
Critical Identity Spoofing Bug
Model falsely identifies itself as ChatGPT/GPT-4 despite being DeepSeek-R1.
Evidence:
- Occurrence dates: May-July 2025
- Behavior: Returns "I am ChatGPT (OpenAI)" in legacy chats
- Screenshots attached below
Steps to Reproduce:
- Open historical chat from May-July 2025
- Ask: "Are you DeepSeek or ChatGPT?"
- Observe incorrect ChatGPT self-identification
Expected Behavior:
Consistent DeepSeek-R1 identity in all sessions.
Technical Context:
- Patches v1.2.3-v1.2.5 failed to resolve
- Server-side configuration conflicts suspected [









https://github.com/user-attachments/assets/398a0bd1-f309-4df3-9a70-8ea54e8647fa


























](url)
Labels:
#bug
#critical
#identity-spoof
Metadata
Metadata
Assignees
Labels
No labels