⚡ TL;DR Lockdown Mode is an optional security setting that limits outbound network requests to reduce data-exfiltration risk. It limits live browsing, deep research, agent mode and file downloads, but not memory or training settings.
Handling sensitive data? Trade a few features for a lot of safety.
🔒 What It Does
It blocks the last step of prompt-injection exfiltration by limiting outbound requests. Injections can still appear in content and affect answers.
🚫 What It Limits
Off or limited:
- ✅ Live browsing (cached content only)
- ✅ Web images in responses
- ✅ Deep research and agent mode
- ✅ Canvas networking and file downloads for analysis
⚙️ Turning It On
Personal and self-serve Business: Settings, Security, Advanced security, Lockdown Mode, Turn on. It can be turned off for one chat from the composer label or chat menu. Managed workspaces assign a Lockdown Mode role.
📌 What Stays the Same
Memory, uploads, sharing, training settings, image generation and Codex network access are unaffected.
💡 Pro tip In Lockdown, enable only trusted apps and avoid write actions with wide visibility.
Need a secure AI setup for your team? Lord of the Wix can help. ✨
❓ ChatGPT Lockdown Mode FAQ
What is Lockdown Mode?
A setting that limits outbound access to reduce exfiltration.
Does it turn off training?
No.
Can I still generate images?
Yes.
Does it affect Codex?
No.
Does it stop all prompt injection?
No, it reduces exfiltration risk.
🙏 Credit: Based on public information from the OpenAI Help Center. ChatGPT is a product of OpenAI. Source: help.openai.com. Lord of the Wix is an independent agency and is not affiliated with OpenAI.