ChatGPT gets Lockdown Mode and risk warnings for safer use: What they do

OpenAI has added two new security features to OpenAI’s ChatGPT platform – Lockdown Mode and “Elevated Risk” labels. The update focuses on reducing the chances of prompt injection attacks and giving users clearer visibility into tools that may carry additional security risks. The company shared details about the changes in an official announcement. Also Read: Using ChatGPT? Here’s What You Should Never Share

Why OpenAI is adding these controls

As ChatGPT increasingly connects to the web and third-party apps, the security landscape has changed. One growing concern is prompt injection, where attackers attempt to trick an AI system into revealing sensitive data or following malicious instructions. Also Read: ‘Patil Effect’ and now Ghose: How Anthropic is building an Indian powerhouse to disrupt Big Tech

According to OpenAI, the new measures are meant to help users, especially those working with sensitive information, manage these risks without removing useful features entirely.

What Lockdown Mode does

Lockdown Mode is an optional security setting aimed at higher-risk users, such as executives or security teams in large organisations. It adds stricter limits on how ChatGPT interacts with external systems.

When Lockdown Mode is turned on, some tools and features are limited or completely disabled to reduce the chances of sensitive data being exposed. For example, browsing is handled differently in Lockdown Mode. Instead of connecting to the live internet, ChatGPT only works with cached content, so no direct external requests are made.

If a feature cannot be made secure with clear guarantees, it may be turned off entirely in this mode. Workspace admins can enable Lockdown Mode through the settings and apply it to selected roles within their organisation. They also have the option to control which connected apps and actions remain available when the mode is active.

At the moment, Lockdown Mode is available for ChatGPT Enterprise, ChatGPT Edu, ChatGPT for Healthcare, and ChatGPT for Teachers. OpenAI has said it plans to make the feature available to individual users in the coming months.

“Elevated Risk” labels explained

Along with this, OpenAI is adding “Elevated Risk” labels across ChatGPT, ChatGPT Atlas, and Codex. The new labels are meant to indicate features that require extra network access or may involve a higher level of data exposure.

For example, in Codex, users can choose to give the assistant internet access so it can look up documentation or interact with websites. When such access is enabled, the settings screen will display an Elevated Risk label explaining what changes and what risks are involved.

Add Techlusive as a Preferred Source

OpenAI says these labels may change over time. As protections improve, some features may no longer carry the warning if the associated risks are reduced.