OpenAI has developed a new security feature to protect its users against command injection attacks, one of the most complex types of cyber attacks. This new layer, called Lockdown Mode, focuses especially on blocking malicious chatbot instructions hidden on web pages or various content sources.

This new security measure aims to help ChatGPT users process their data in a more secure environment. The company states that this feature is designed especially for people and institutions working with sensitive data.

Restrictions that come with Lockdown Mode

When Lockdown Mode is activated, significant changes occur in the operation of ChatGPT. Within the scope of this mode, the live web browsing feature is disabled and users can only access cached content.

Additionally, image retrieval and display functions over the internet cannot be used in this mode. However, users can continue to create their own images.

The restrictions in the system do not end there. Advanced features such as deep research and spy mode also become unavailable when Lockdown Mode is on.

Security breaks and access details

OpenAI emphasizes that ChatGPT is not completely immune to attacks, even if Lockdown Mode is active. The company warns that commands contained in cached web content or uploaded documents may continue to affect the accuracy or behavior of the answers.

The main purpose of this feature is to minimize the risks of data leakage and reduce the possibility of sharing sensitive information. OpenAI states that this mode is not suitable for everyone and has been developed especially for users who give high priority to data security.

The new security feature is currently being rolled out gradually. OpenAI has begun rolling out Lockdown Mode support to self-service ChatGPT Business accounts and eligible personal accounts.

Do you think this new security measure will be enough to reduce the risks of data leakage on ChatGPT?