OpenAI Expands ChatGPT Lockdown Mode to All Users

OpenAI expanded Lockdown Mode, an optional security setting for ChatGPT, to all account tiers including Free, Go, Plus, Pro, and self-serve Business plans. Previously available only to Enterprise, Education, Healthcare, and Teacher subscribers since February, the feature is designed to protect users and organizations handling sensitive data from prompt injection attacks, which use hidden instructions in webpages or documents to trick AI into exfiltrating data.

To mitigate these risks, Lockdown Mode restricts outbound network requests and prevents the model from transmitting data to external sources. Enabling the mode requires significant functional trade-offs, as it disables Deep Research, Agent Mode, Developer Mode, and network-enabled Canvas functions. It also blocks live web browsing—forcing reliance on cached content—and prevents the system from downloading files for analysis or pulling images from the internet. Users can still generate images and manually upload documents.

OpenAI clarified that the mode is not a complete defense, as prompt injections can still appear in cached content or uploaded files. However, the company stated the goal is to reduce the likelihood of sensitive data being shared. Alongside this rollout, OpenAI introduced an active session manager allowing users to monitor devices accessing their accounts and remotely log out of sessions.