Lockdown Mode was reportedly designed to help protect ChatGPT against prompt injection attacks. The feature is intended for users handling sensitive information who want additional security protections.
OpenAI confirmed Lockdown mode offers an additional level of security along with existing options that ChatGPT can offer. In a prompt injection attack, the hackers manage to insert some secret commands within the text that is processed by AI.
Lockdown Mode restricts certain ChatGPT features to reduce the risk of exposing sensitive information. ChatGPT will be unable to download images from the web or display images in response to queries. File uploads are permitted for review, but ChatGPT cannot access files available on the internet in Lockdown Mode.
Deep Research and Agent Mode are not available in Lockdown Mode since they require access to the internet. When OpenAI announced the introduction of Lockdown Mode, they explained that the tool would not prevent the occurrence of any unsafe prompts. Instead, it aims to make it harder for an attacker to gain access to user information.
This feature release follows ongoing warnings from security researchers concerning prompt injection attacks, which are considered a major risk for AI technologies. Prompt injections are when secret instructions are inserted into websites, PDF documents, and even other types of files that the AI models process.
These injections try to make the AI system reveal confidential information and take unauthorized action on behalf of the user. Lockdown Mode would reportedly block connections to third-party services to reduce the risk of data exposure. However, some functions become less effective since the AI is deprived of access to up-to-date information from the Internet.
Another addition is the Active Session Manager, which provides users with the ability to know the current state of active sessions using their ChatGPT account from various devices and browsers. Users can log off from a particular device or all current sessions, depending on whether the situation requires doing so as there seems to be any unusual activity.
Anyone who uses ChatGPT will have access to Lockdown Mode, including even free users. The feature can be activated under the Safety & Security tab in your settings menu. The development reflects the growing focus on security as AI tools increasingly handle sensitive information.
Also Read: OpenAI Accused of Sharing ChatGPT User Data With Google and Meta