OpenAI rolled out Lockdown Mode, an optional security setting that adds a hard‑line defense against prompt injection attacks. Prompt injection—malicious instructions hidden in web content that trick chatbots—has risen as models pull more internet data. The new mode targets users who handle sensitive information, offering stricter safeguards without requiring a paid plan.

When enabled, Lockdown Mode trims several ChatGPT capabilities. Image generation remains, but the model cannot fetch external pictures or embed them in replies. File downloads are blocked, though manual uploads still work. Features such as Deep Research and Agent Mode are disabled entirely. Admins can enforce it organization‑wide for consistency. The setting does not alter memory, conversation sharing, or model‑training usage, all of which stay configurable.

The toggle lives in ChatGPT’s Settings under Safety and security → Advanced security. Users can switch it on per chat and turn it off temporarily. Alongside Lockdown Mode, OpenAI introduced an active session manager that lists devices and browsers accessing the account, letting users log out individually or all at once—a process that may take up to thirty minutes. Enterprises gain clearer oversight of account activity.