According to OpenAI, the ChatGPT Lockdown Mode—previously limited to enterprise and education users—has been expanded to all personal and commercial accounts across Free, Plus, Pro, and ChatGPT Business plans.
The security feature disables multiple network and analysis functions to prevent prompt injection attacks. In Lockdown Mode, real-time web browsing is restricted to cached content only, while Deep Research and Agent Mode are completely disabled. Code generated in Canvas cannot access the network, and file download channels for data analysis are severed. Image retrieval from the web is also restricted, though users can still generate and upload images. According to OpenAI, Lockdown Mode does not prevent prompt injection from affecting model logic or response accuracy, but prevents sensitive data exfiltration.