Bug: AI consistently outputs garbled text (乱码) for Chinese content during code generation and editing

What version of the Codex App are you using (From “About Codex” dialog)?

26.325

What subscription do you have?

Pro

What platform is your computer?

Microsoft Windows NT 10.0.19045.0 x64

What issue are you seeing?

We have been forced to add explicit rules in our workspace instructions (e.g., "Treat source code, config files, docs, and SQL files as UTF-8 text", "Never introduce garbled text") in an attempt to prevent this, but the issue still sporadically occurs. We suspect there is a text encoding handling issue when Codex processes UTF-8 multibyte characters. A native fix to ensure reliable UTF-8 handling is highly needed.

What steps can reproduce the bug?

When using Codex to generate, edit, or refactor files (such as source code, config files, docs, and SQL files) that contain Chinese characters, the output frequently breaks the file encoding and results in garbled text (乱码). This issue has occurred multiple times and severely impacts the development experience for projects that require Chinese comments or content.

What is the expected behavior?

_No response_

Additional information

_No response_