OpenAI has rolled out two new security features for ChatGPT users: Lockdown Mode and “Elevated Risk” Flags. Together, these tools provide users with clearer insights into potential risks and greater control over their interactions with the system.

Lockdown Mode is an optional, advanced security feature designed for a limited group of users who face a higher risk of cyberattacks.

This includes executives, security experts, and teams at large organizations. It’s not intended for everyday use, but for those who require an extra layer of protection.

When activated, Lockdown Mode strictly restricts how ChatGPT interacts with external systems. It disables specific tools and features that could be exploited by attackers using prompt injection, aiming to prevent sensitive data from being accessed through hidden or malicious instructions.

For instance, in Lockdown Mode, web browsing is limited to cached content, so no live network requests leave OpenAI’s secure environment.

Any feature that cannot ensure strong data protection may be completely disabled.

Workspace administrators can activate Lockdown Mode by creating a specific role in the workspace settings. Enabling it applies extra restrictions in addition to the current security settings.