en.Wedoany.com Reported - US artificial intelligence company OpenAI has launched a new security feature called "Block Mode," aimed at enhancing protection for users and organizations that handle sensitive information while using large language models.

This mode is designed to detect and defend against attempts to leak sensitive data to unauthorized parties through prompt injection attacks. OpenAI stated that Block Mode is not available to all users, but is intended for individuals and organizations that handle sensitive data and seek stricter protection against data leakage risks associated with prompt injection.

Prompt injection attacks are a new method employed by hackers, where attackers can hide instructions, such as code documents, in web pages and upload them to large language models. While processing this content, attackers can manipulate the AI system to obtain sensitive information. OpenAI chose to launch this mode amid the rising trend of such attacks.

Once Block Mode is enabled, ChatGPT will restrict several large language model functions, including not fetching internet images or generating images, and limiting users' ability to download files for analysis. However, users can still upload files, images, and documents for analysis. OpenAI clarified that this mode will not affect conversation memory or conversation sharing settings. In enterprise environments, administrators will have full configuration and change permissions.

Additionally, ChatGPT has introduced a new tool called "Active Session Manager," allowing users to view logged-in devices and even browsers, and remotely log out of accounts from unwanted devices.

This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com