Enable Lockdown Mode for stricter ChatGPT security
Optional setting to limit or turn off web and external service connections for enhanced security.
Lockdown Mode
Settings > Security (Personal) or Workspace Settings (Business/Enterprise)
Reduces ways prompt-injection attacks could cause information to be shared outside the conversation by limiting or disabling features that connect ChatGPT to the web or external services.
Lockdown Mode is an optional security setting for ChatGPT users who need stricter guardrails when working with sensitive information. It limits or turns off features like live web access, image support, and connectors to reduce the risk of data exfiltration via prompt injection attacks.
By disabling or restricting features that interact with external systems, Lockdown Mode reduces the attack surface for prompt injection, preventing malicious instructions from causing sensitive data to be shared.
- Introducing Lockdown Mode and Elevated Risk labels in ChatGPT ↗OpenAI · OpenAI News · Feb 13, 2026 · paraphrased
First seen on Subvolts Feb 13, 2026. Updated Sep 11, 2026. Spotted a problem? Tell us.
- Introducing Lockdown Mode and Elevated Risk labels in ChatGPTOpenAI · OpenAI News · Feb 13, 2026
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- Safety tipSafeguard sensitive data with ChatGPT agentExercise caution when using ChatGPT agent with sensitive data and logins. · 5 min
- Safety tipDefend against prompt injection attacks with careful oversightPrevent malicious instructions in third-party websites from misleading AI models. · 5 min
- ChecklistChecklist for Spotting Scam Messages with ChatGPTUse this checklist to identify common warning signs in suspicious emails or texts with ChatGPT. · 1 min
- ChecklistChecklist for applying AI to daily work tasksSteps to effectively use AI for improving workflows, creating repeatable processes, and directing agents. · 15 min
- ChecklistChecklist for students using AI for academic tasksSteps for students to use AI for organizing study plans, group projects, assignments, and career preparation. · 30 min
- HeuristicUse ChatGPT as a Second Opinion for Scam MessagesLeverage ChatGPT to help identify warning signs in suspicious messages. · 2 min
Questions people ask
- Which AI tools does this work with?
- The sources used it with ChatGPT.
- Where did this come from?
- OpenAI (OpenAI News). Subvolts collected, normalized and credited it; the link above opens the original.