OpenAI has begun rolling out an optional Lockdown Mode to provide extra security against prompt injection attacks for users handling sensitive data. The feature limits certain capabilities in ChatGPT and other products to reduce risks of data exfiltration. It is available to all personal accounts, including those on the free tier.
OpenAI describes Lockdown Mode as a last line of defense, building on existing protections in its models and systems. The company states that most users do not need the feature, which is intended for people and organizations that require stricter safeguards. Enabling the mode restricts several functions. Users can still generate images and upload photos but cannot pull images from the internet or display them in responses. The chatbot cannot download files for analysis, and features like Deep Research and Agent Mode are disabled entirely. Lockdown Mode does not alter memory, file uploads, conversation sharing, or the use of conversations to improve models. OpenAI is also introducing an active session manager that lets users view and log out of devices or browsers accessing their accounts, with full logouts taking up to 30 minutes. To activate the setting, users open ChatGPT settings, select Safety and security, then Advanced security, and toggle on Lockdown Mode. Temporary deactivation is available per chat.
