Bug Report: Memory Functionality Automatically Converts Japanese Entries to English, Removing First-Person Perspective and Hindering Chronological Experiment Tracking

ChatGPT Bugs
1

When attempting to accurately register memories in Japanese using the first-person pronouns (“僕” or “私”) along with specific dates, the system automatically translates these entries into English and alters the original text, removing essential first-person perspective and changing the nuances entirely.

This is not just a linguistic issue. It severely impacts the ability to maintain accurate, chronological records of experimental notes. Furthermore, by forcibly translating into English, users who are not proficient in English are unfairly disadvantaged. For a system that claims multilingual support, this feels distinctly unfriendly—even discriminatory—towards Japanese users.

Steps to Reproduce:

  1. Attempt to register a memory exactly as written in Japanese first-person format (e.g., “2025年6月3日、僕は〇〇を実施した。”).
  2. Check the actual registered memory.
    → It appears as something like “June 3, 2025, Conducted 〇〇,” losing the critical first-person pronouns (“僕” or “私”) and other linguistic nuances.

Expected Behavior:

Memories should be registered exactly as originally entered in Japanese, without translation or modification. Specifically, first-person pronouns are crucial for maintaining the intended context and meaning, and should never be altered or removed.

I am using ChatGPT extensively for research and development purposes, specifically leveraging the memory feature for logging critical experiment data in Japanese. However, when I attempt to save logs verbatim with precise dates and first-person perspective in Japanese, the memory feature arbitrarily modifies my inputs by translating them into English and stripping the first-person perspective. This behavior makes it impossible to maintain accurate chronological records and severely impairs the continuity required for my research.

Since this issue directly impacts development and experimental workflows, I believe it falls within the scope of the developer community forum. Could you please provide guidance or escalate this as a technical issue?

Prompt action on this issue would be greatly appreciated.

2

This issue is critical to my ongoing research and directly related to ChatGPT’s core functionalities. I’d greatly appreciate if you could at least escalate this internally or provide specific guidance on the appropriate contact or channel.