ChatGPT Lockdown Mode is an optional security setting from OpenAI that restricts ChatGPT's network capabilities to reduce the risk of data exfiltration from prompt injection attacks. If you handle sensitive data in ChatGPT (customer info, contracts, financial data), this guide covers whether you should turn it on, how to set it up, and what you'll lose by doing so.
This article is based on OpenAI's official help documentation and TechCrunch's coverage, verified as of June 8, 2026.
What Lockdown Mode Does
Lockdown Mode launched on June 5–6, 2026 as an optional security setting for people handling sensitive data. It reduces the risk of data exfiltration from prompt injection attacks.
Simply put: it limits ChatGPT's ability to make network requests, so malicious instructions hidden in web pages or files can't siphon your data out. Lockdown Mode won't prevent prompt injections from appearing in ChatGPT's responses, but it reduces the paths attackers can use to extract data from your account.
What You Lose with Lockdown Mode
Fully disabled
- Deep Research
- Agent Mode
- Canvas networking—code generated in ChatGPT Canvas can't access the network
Partially limited
- Web browsing: only cached content, no live browsing
- Web images: ChatGPT won't fetch or display images from the web (you can still generate images)
- File downloads: ChatGPT can't download files for analysis (you can still upload files manually)
Unaffected
- Memory
- Manual file uploads
- Conversation sharing
- Codex network access (Lockdown Mode doesn't affect Codex)
Note: Lockdown Mode and Developer Mode can't be used at the same time. Turning on Lockdown Mode disables Developer Mode, and vice versa.
Who Should Turn It On (and Who Shouldn't)
Lockdown Mode isn't for everyone. Here's what OpenAI says: "It is designed for people and organizations that handle sensitive data and want stricter protection from data exfiltration risks related to prompt injection."
Consider turning it on if:
- You include customer personal info, financial data, medical information, or trade secrets in ChatGPT conversations
- Your organization uses ChatGPT for contracts, legal documents, or internal policies
- Your team has compliance requirements (SOC 2, HIPAA reference scenarios)
Skip it if:
- You use ChatGPT for everyday chat, learning, or creative writing with no sensitive data
- You rely on Deep Research or Agent Mode regularly
- Your workflow depends on live web browsing
Setup Steps
Lockdown Mode is rolling out to all account types, including Free, Plus, Pro, and self-serve ChatGPT Business. If you don't see the option yet, check back later.
- Open ChatGPT Settings
- Go to "Security"
- Under "Advanced security," find "Lockdown mode"
- Toggle it on
Once enabled, a status indicator appears above the chat composer showing that Lockdown Mode is active.
Turning It Off for a Single Chat
If you need full features for one conversation:
- In a Lockdown Mode session, click the status indicator above the chat composer
- Select "Manage"
- Choose "Turn off for this chat"
This only affects the current conversation, not your other chats.
To turn it back on, open the same chat's menu (•••), select Lockdown > Enabled.
What Lockdown Mode Can't Do
- It can't fully prevent prompt injection—malicious instructions can still appear in cached content or uploaded files and affect response accuracy.
- It's not a security patch—it just restricts network requests, not model vulnerabilities.
- There's a feature trade-off—Deep Research and Agent Mode are unavailable. Evaluate whether your workflow depends on them before turning it on.
Post-Setup Checklist
- Confirm the Lockdown Mode status indicator appears above the chat composer
- Test web browsing to verify only cached content is returned
- Test Deep Research and Agent Mode to confirm they're disabled
- If you need Developer Mode, make sure Lockdown Mode is off (they can't coexist)
- Review whether your daily workflow depends on any disabled features before leaving it on long-term