OpenAI has announced the incorporation of a new blocking mode in ChatGPT, officially presented on February 13. This measure affects users considered high-risk, such as journalists, activists, and professionals who carry out their work in sensitive environments.
Advanced configuration for vulnerable collectives
Lockdown mode is an advanced and voluntary security option. When activated, ChatGPT's web browsing is limited to the exclusive use of cached content, without access to real-time information from the Internet. The deep research function and agent mode, which allows autonomous actions, remain deactivated while this system is active.
This setting also prevents ChatGPT from including images in its responses, although users retain the ability to upload photographs for analysis. Additionally, file downloading for analysis is blocked under this mode.
Measures to reinforce protection
The main objective of this lockdown mode is to mitigate prompt injection attacks, a technique that can compromise data security or the integrity of responses generated by artificial intelligence.
Currently, lockdown mode is available for users of ChatGPT Enterprise, ChatGPT Edu, ChatGPT Healthcare, and ChatGPT Teachers. OpenAI has communicated that the feature will be extended to domestic consumer plans in the coming months.
- The lockdown mode is optional and can be activated according to the user's needs.
- Navigation is limited to information previously stored in cache.
- Deep research and agent mode functions are not available under this system.
- File download and the inclusion of images in the responses are restricted.
The company maintains that these measures seek to offer greater protection to groups exposed to specific risks in the use of artificial intelligence tools.