ChatGPT Expands Lockdown Mode to All Users, Blocking Live Web Tools to Curb Data Theft
Updated
Updated · ZDNet · Jun 8
ChatGPT Expands Lockdown Mode to All Users, Blocking Live Web Tools to Curb Data Theft
3 articles · Updated · ZDNet · Jun 8
Summary
OpenAI is rolling out ChatGPT Lockdown mode across Free, Go, Plus, Pro and Business plans after first launching it in February for Enterprise, Edu, Healthcare and Teachers users.
The setting is designed to reduce data theft from prompt-injection attacks by blocking outbound requests to the live internet and external file services, keeping sensitive information from being pulled out of chats.
Lockdown mode does not stop prompt injection itself and still leaves exposure to malicious commands tied to cached web content or uploaded files.
The tradeoff is broad feature loss: live web browsing, web images, deep research, agent mode, Canvas network access and file downloads are disabled, though users can still upload files and images.
OpenAI positions the option for people handling sensitive or confidential data, but the phased rollout means it may not yet appear on every account.