ChatGPT gets a Lockdown Mode: What it does and who actually needs it

OpenAI has begun rolling out Lockdown Mode for personal ChatGPT accounts, an optional security setting built to blunt prompt injection attacks. First introduced for enterprise plans in February, the feature is now reaching Free, Go, Plus, Pro, and self-serve Business users through Settings under Safety and security.

Prompt injection is social engineering aimed at chatbots. Attackers plant malicious instructions inside webpages, documents, or other content an AI might scan, hoping the model obeys them instead of the user. OpenAI is positioning Lockdown Mode as a last line of defence rather than a fix. The company says the mode will not stop prompt injections from appearing in content ChatGPT processes—it just makes it harder for an attacker to walk away with anything sensitive.

The trade-off is aimed at users who handle data worth stealing

Switching it on tightens what ChatGPT can do with the outside world. Live web browsing is restricted to cached content, Deep Research and Agent Mode are switched off entirely, file downloads are blocked, and the chatbot stops pulling or displaying web images inside responses. Image generation still works. Manual file uploads still work. Memory stays untouched. Users can also turn Lockdown off for a single chat through the status banner above the composer when a task needs the full feature set.

A second security tool is arriving quietly alongside it

OpenAI is pairing the launch with an active session manager that lists every device and browser signed into your account. Users can log out individually or in one shot, though the bulk action can take up to 30 minutes to push through. The company is clear that Lockdown Mode is not meant for everyone. It is built for executives, security teams, and anyone whose ChatGPT conversations carry information worth protecting.