OpenAI just added a new ChatGPT security mode — but it disables some popular features

Akarsh Rasik
4 Min Read
Highlights
  • OpenAI has launched Lockdown Mode, a new optional security feature for ChatGPT and other OpenAI products.
  • The feature helps reduce data exfiltration risks by limiting internet access and outbound network requests.
  • Lockdown Mode is available to all users and organizations and is designed for those handling sensitive or confidential data.

OpenAI has started rolling out Lockdown Mode, a new security feature for ChatGPT and other OpenAI products. The optional setting is designed to help protect users from prompt injection attacks, a type of attack that can try to trick AI systems into revealing sensitive information. The feature is available for all OpenAI account types and workspaces. Users must be signed in to enable it.

What is Lockdown Mode?

Lockdown Mode is an advanced security setting that limits some ChatGPT features that connect to the internet or external services.

According to OpenAI, the goal of the feature is to reduce the risk of sensitive data being sent to attackers through prompt injection attacks. The company says it is mainly intended for people and organizations that work with confidential or sensitive information.

How it helps protect data

OpenAI explained that prompt injection remains a difficult security challenge across the AI industry. The security feature adds another layer of protection by restricting outbound network requests. This can help stop attackers from using ChatGPT to send sensitive information outside the platform.

However, OpenAI noted that the setting does not stop malicious prompts from appearing in web content or uploaded files. Such prompts may still affect how ChatGPT responds.

Features disabled in Lockdown Mode

When the feature is turned on, several capabilities become unavailable or limited.

Live web browsing

ChatGPT can only access cached web content instead of browsing the live internet. This means search results may be limited or outdated.

Deep Research

The Deep Research feature is completely disabled.

Agent Mode

Users cannot use Agent Mode while Lockdown Mode is enabled.

Image retrieval

ChatGPT cannot fetch images from the web or show images in regular responses. Users can still upload images and use image generation features where available.

Network access in Canvas

Canvas-generated code cannot connect to external networks.

File downloads

ChatGPT cannot download files for analysis. Users can still upload files manually.

What stays the same?

OpenAI said Lockdown Mode does not change several existing settings. Features such as memory, file uploads, conversation sharing, and model training preferences remain unchanged. These settings can still be managed separately. The company also confirmed that Lockdown Mode does not affect network access in Codex.

Changes to apps and connectors

For personal accounts and self-serve ChatGPT Business accounts, Lockdown Mode allows access to synced connector data but blocks live connector access and write actions. Some services, including finance-related experiences and shopping-agent features, are not available while the mode is active.

For enterprise and managed workspaces, administrators can decide which apps and connectors users can access through role-based permissions.

Guidance for businesses

OpenAI recommends that organizations using Lockdown Mode only allow trusted apps and actions. The company warns that write actions in apps can create a higher risk because they can send information outside the system. Read-only access is generally considered safer.

Enterprise customers can also use OpenAI’s Compliance API Logs platform to monitor app activity and data access.

How to turn on Lockdown Mode

Personal users and self-serve ChatGPT Business customers can enable the feature by following these steps:

  1. Open Settings
  2. Select Security
  3. Go to Advanced Security
  4. Turn on Lockdown Mode
  5. Confirm the change

OpenAI said Lockdown Mode and Developer Mode cannot be used at the same time. Enabling one will automatically disable the other. For managed workspaces, administrators can create special Lockdown Mode roles and assign them to users or groups.

Keep up with the tech that actually matters.

From breaking news to deep dives, TrueTech brings you the tech stories worth knowing.
Add us as a preferred source on Google Search for quicker access to our coverage.

Add TrueTech as a preferred source on Google
Share This Article
Follow:
Writing about emerging gadgets and technology news, as well as keeping you updated on movie and music news, with a focus on all things K-pop.
Highlights
  • OpenAI has launched Lockdown Mode, a new optional security feature for ChatGPT and other OpenAI products.
  • The feature helps reduce data exfiltration risks by limiting internet access and outbound network requests.
  • Lockdown Mode is available to all users and organizations and is designed for those handling sensitive or confidential data.

OpenAI has started rolling out Lockdown Mode, a new security feature for ChatGPT and other OpenAI products. The optional setting is designed to help protect users from prompt injection attacks, a type of attack that can try to trick AI systems into revealing sensitive information. The feature is available for all OpenAI account types and workspaces. Users must be signed in to enable it.

What is Lockdown Mode?

Lockdown Mode is an advanced security setting that limits some ChatGPT features that connect to the internet or external services.

According to OpenAI, the goal of the feature is to reduce the risk of sensitive data being sent to attackers through prompt injection attacks. The company says it is mainly intended for people and organizations that work with confidential or sensitive information.

How it helps protect data

OpenAI explained that prompt injection remains a difficult security challenge across the AI industry. The security feature adds another layer of protection by restricting outbound network requests. This can help stop attackers from using ChatGPT to send sensitive information outside the platform.

However, OpenAI noted that the setting does not stop malicious prompts from appearing in web content or uploaded files. Such prompts may still affect how ChatGPT responds.

Features disabled in Lockdown Mode

When the feature is turned on, several capabilities become unavailable or limited.

Live web browsing

ChatGPT can only access cached web content instead of browsing the live internet. This means search results may be limited or outdated.

Deep Research

The Deep Research feature is completely disabled.

Agent Mode

Users cannot use Agent Mode while Lockdown Mode is enabled.

Image retrieval

ChatGPT cannot fetch images from the web or show images in regular responses. Users can still upload images and use image generation features where available.

Network access in Canvas

Canvas-generated code cannot connect to external networks.

File downloads

ChatGPT cannot download files for analysis. Users can still upload files manually.

What stays the same?

OpenAI said Lockdown Mode does not change several existing settings. Features such as memory, file uploads, conversation sharing, and model training preferences remain unchanged. These settings can still be managed separately. The company also confirmed that Lockdown Mode does not affect network access in Codex.

Changes to apps and connectors

For personal accounts and self-serve ChatGPT Business accounts, Lockdown Mode allows access to synced connector data but blocks live connector access and write actions. Some services, including finance-related experiences and shopping-agent features, are not available while the mode is active.

For enterprise and managed workspaces, administrators can decide which apps and connectors users can access through role-based permissions.

Guidance for businesses

OpenAI recommends that organizations using Lockdown Mode only allow trusted apps and actions. The company warns that write actions in apps can create a higher risk because they can send information outside the system. Read-only access is generally considered safer.

Enterprise customers can also use OpenAI’s Compliance API Logs platform to monitor app activity and data access.

How to turn on Lockdown Mode

Personal users and self-serve ChatGPT Business customers can enable the feature by following these steps:

  1. Open Settings
  2. Select Security
  3. Go to Advanced Security
  4. Turn on Lockdown Mode
  5. Confirm the change

OpenAI said Lockdown Mode and Developer Mode cannot be used at the same time. Enabling one will automatically disable the other. For managed workspaces, administrators can create special Lockdown Mode roles and assign them to users or groups.

Keep up with the tech that actually matters.

From breaking news to deep dives, TrueTech brings you the tech stories worth knowing.
Add us as a preferred source on Google Search for quicker access to our coverage.

Add TrueTech as a preferred source on Google
Share This Article
Follow:
Writing about emerging gadgets and technology news, as well as keeping you updated on movie and music news, with a focus on all things K-pop.