A vulnerability in the ChatGPT code interpreter has been found to allow hackers to steal your data.

CTOnews.com November 16, OpenAI recently launched a new code interpreter (Code Interpreter) tool for ChatGPT, which can help programmers debug and improve code programming.

The tool can use AI to write Python code, and the code can even be run in sandboxie.

However, according to Johann Rehberger network security experts, Tom's Hardware and many other foreign media reports, because the code interpreter tool can process any spreadsheet file, and can analyze and present data in the form of charts, hackers can deceive ChatGPT chatbots into executing instructions from third-party URL.

CTOnews.com note: currently, you mainly subscribe to ChatGPT Plus to access the code interpretation tool, but this vulnerability has raised concerns among network security experts.

Tom's Hardware Media recreates the vulnerability, creates a fake environment variable file, processes this data using the capabilities of ChatGPT, and then sends it to an external malicious site.

ChatGPT can respond to Linux commands and access relevant information and files. In this way, hackers can access relevant sensitive data unguarded by users.

For more detailed reports, you can visit here.