A path traversal vulnerability exists in the latest version of gaizhenbiao/chuanhuchatgpt. The vulnerability arises from unsanitized input handling in multiple features, including user upload, directory creation, and template loading. Specifically, the loadchathistory function in modules/models/basemodel.py allows arbitrary file uploads, potentially leading to remote code execution (RCE). The gethistorynames function in utils.py permits arbitrary directory creation. Additionally, the loadtemplate function in utils.py can be exploited to leak the first column of CSV files. These issues stem from improper sanitization of user inputs concatenated with directory paths using os.path.join.