OpenAI opens locked mode to all ChatGPT personal users to limit web access and reduce data leak risks

OpenAI has opened its locked mode to all ChatGPT personal users, including Free, Go, Plus, and Pro tiers. Previously only for enterprise users and security teams, the mode aims to reduce data leak risks by limiting features that require the internet or external services.

When enabled, ChatGPT restricts live web browsing, deep research, and Agent mode. Web browsing only accesses cached content, search results may be limited, and image support might be unavailable—though users can still upload images and generate new ones. File downloads and Agent mode are also disabled, but manually uploaded files remain usable.

OpenAI notes that locked mode can't fully prevent prompt injection attacks, as malicious instructions might sneak in through uploaded files or cached web pages. This feature offers extra data protection for privacy-sensitive work or research scenarios.