OpenAI introduces Lockdown Mode to block data exfiltration from prompt injection attacks

OpenAI has introduced Lockdown Mode, an optional advanced security feature for ChatGPT that restricts outbound network requests to mitigate data exfiltration risks from prompt injection attacks. The mode disables or limits capabilities including live web browsing, deep research, agent mode, and external file downloads, while allowing cached content and manual uploads. It is available for eligible personal, Plus, Pro, and self-serve Business accounts, with workspace admins able to enforce it via role-based access controls. The feature is aimed at users handling sensitive data who need stricter protections, though it does not prevent prompt injections from appearing in processed content.