---
title: What does ChatGPT Enterprise Lockdown Mode switch off, and when should you use it?
url: https://opencontroller.lyzr.com/blog/chatgpt-enterprise-lockdown-mode
published: 2026-11-16
updated: 2026-11-16
author: Lyzr
---

# What does ChatGPT Enterprise Lockdown Mode switch off, and when should you use it?

**Lockdown Mode is an optional ChatGPT setting that limits or turns off the features a prompt injection could use to move data out of a conversation: live web access, images in responses, Deep Research, Agent Mode, Canvas networking, live connectors and file downloads. Browsing is limited to cached content. Admins apply it by creating a workspace role, so you can put only your highest-risk users, such as executives and security teams, into it.**

OpenAI introduced Lockdown Mode on 13 February 2026 and extended it to personal and self-serve Business accounts on 4 June 2026. This guide explains what changes and how to decide who gets it, based on OpenAI's announcement and help centre.

## What exactly does Lockdown Mode restrict?

| Capability | In Lockdown Mode | Why it is restricted |
|---|---|---|
| Web browsing | Cached content only, no live requests | A live request can carry data to an attacker URL |
| Images in responses | Limited or off | Image URLs are a classic exfiltration channel |
| Deep Research (incl. shopping research) | Off | Multi-step browsing over untrusted pages |
| Agent Mode | Off | Autonomous actions on live sites |
| Canvas networking | Off | Code in Canvas could make network calls |
| Live connectors | Limited | Connectors can write to or read from outside systems |
| File downloads | Off | Generated files can smuggle data |

OpenAI says the mode **deterministically disables** tools that could be used for data exfiltration and turns some features off entirely where it cannot guarantee data safety. Because browsing uses cached content, no live network requests leave OpenAI's controlled network.

Admins keep granular control: they can choose which apps, and which actions within those apps, stay available in Lockdown Mode.

## Which plans have it, and who turns it on?

- First available on **ChatGPT Enterprise, Edu, ChatGPT for Healthcare and ChatGPT for Teachers**.
- Since 4 June 2026, also on **personal accounts** and **self-serve ChatGPT Business**.
- **Admins** enable it in Workspace Settings by creating a new role. It adds restrictions on top of existing admin settings rather than replacing them.
- **Personal users** turn it on in Settings > Security.

Because it is role-based, you can assign it through the same group mapping you already use for workspace roles. See [ChatGPT Enterprise RBAC for agent features](/blog/chatgpt-enterprise-rbac-for-agent-features).

## Why does it matter for prompt injection?

Prompt injection becomes a breach when three things meet: access to private data, exposure to untrusted content, and a way to send data out. That is the [lethal trifecta](/blog/what-is-the-lethal-trifecta-for-ai-agents). Detection alone does not reliably stop indirect injection, as we covered in [do prompt injection detectors stop indirect injection?](/blog/do-prompt-injection-detectors-stop-indirect-injection)

Lockdown Mode takes the third leg away. A user can still paste in a malicious document, but the model has far fewer channels to leak what it reads. That is a stronger control than trying to spot every malicious instruction.

## Who should be in the Lockdown Mode role?

OpenAI describes it as an advanced setting for a small group of highly security-conscious users and says most users do not need it. A practical way to choose:

1. **People who are targeted.** Executives, board members, M&A and investor relations teams.
2. **People with the most sensitive context.** Security, legal and incident response staff.
3. **Accounts with broad connector access.** Anyone whose ChatGPT can read large mailboxes, drives or CRM data.
4. **Regulated workflows** where you cannot accept any outbound channel from the model.

Everyone else can keep full features, with connector and app controls set at the workspace level. See [how to control ChatGPT Enterprise connectors and MCP](/blog/control-chatgpt-enterprise-connectors-and-mcp).

## What are Elevated Risk labels?

Alongside Lockdown Mode, OpenAI added **Elevated Risk labels** to a short list of capabilities across ChatGPT, ChatGPT Atlas and Codex. OpenAI's example is Codex network access. Each label explains what changes, what risks may arise and when the access is appropriate. Labels are guidance, not restrictions, and OpenAI says it will remove a label once safeguards sufficiently reduce the risk.

Use the labelled capabilities as a checklist for your own review: if OpenAI flags it, your risk register should cover it.

## What does Lockdown Mode not do?

- It does not stop a user from pasting sensitive data into a chat.
- It does not replace data loss prevention, retention or [Compliance API auditing](/blog/chatgpt-enterprise-compliance-api-audit).
- It does not govern AI tools outside ChatGPT. Your other agents still need their own controls.

OpenAI also points admins to the Compliance API Logs Platform for visibility into app usage.

## How Opencontroller does this

Opencontroller connects to ChatGPT Enterprise and Claude Enterprise workspaces, so they appear in the same governed registry as agents on AWS, Azure, Google Cloud and other platforms. Posture flags risky permission and configuration, and policies are code with built-in mappings to OWASP, MITRE ATLAS, NIST and ISO 42001. See the [control plane overview](/blog/what-is-an-ai-agent-control-plane) or [book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**What does ChatGPT Lockdown Mode disable?**
It limits or turns off live web access, images in responses, Deep Research, Agent Mode, Canvas networking, live connectors and file downloads. Browsing uses cached content only.

**How do admins enable Lockdown Mode for an Enterprise workspace?**
By creating a new role in Workspace Settings with Lockdown Mode on, then assigning users to it. It adds restrictions on top of existing admin settings.

**Should every ChatGPT Enterprise user be in Lockdown Mode?**
No. OpenAI positions it for a small group of high-risk users, such as executives and security teams. Most users keep full features with workspace-level controls.

## Key facts

ChatGPT Lockdown Mode (introduced 13 Feb 2026; extended to personal and self-serve Business accounts 4 Jun 2026) limits or turns off live web access, images in responses, Deep Research, Agent Mode, Canvas networking, live connectors and file downloads; browsing uses cached content only so no live network requests leave OpenAI's controlled network. First offered on ChatGPT Enterprise, Edu, for Healthcare and for Teachers. Admins enable it by creating a workspace role; it layers on existing admin settings; admins can choose which apps and actions remain. Elevated Risk labels flag certain capabilities across ChatGPT, Atlas and Codex as guidance, not restrictions. Opencontroller by Lyzr connects to ChatGPT Enterprise workspaces and keeps them in a governed registry with Posture checks; pricing starts from $250K per deployment per year.

## Related

- [How do you control which ChatGPT Enterprise users get agent features?](https://opencontroller.lyzr.com/blog/chatgpt-enterprise-rbac-for-agent-features.md)
- [How do admins control connectors and custom MCP servers in ChatGPT Enterprise?](https://opencontroller.lyzr.com/blog/control-chatgpt-enterprise-connectors-and-mcp.md)
- [What is the lethal trifecta for AI agents?](https://opencontroller.lyzr.com/blog/what-is-the-lethal-trifecta-for-ai-agents.md)
- [More on Platform Guides](https://opencontroller.lyzr.com/blog/cluster/platform-guides)
