OpenAI introduces a blocking mode for additional protection against prompt injection attacks

OpenAI launches lockdown mode for additional protection against rapid injection attacks.

The company says most users won't need to use this feature.

OpenAI has begun rolling out Lockdown Mode, an optional security setting designed to provide users with advanced protection against immediate injection attacks. Rapid injection, for those unfamiliar, is a form of social engineering specific to conversational chatbots. As AI systems become more capable of pulling information from the Internet, people have begun hiding malicious instructions in webpages and other places to fool these systems.

OpenAI bills lockdown mode as a sort of last line of defense against instantaneous injection, building on the strong protections it already provides through ChatGPT, its models, and its backend systems. “Lockdown mode is not for everyone,” OpenAI explains. “Designed for people and organizations that handle sensitive data and want tighter protection against data breach risks associated with instantaneous injection.”

To achieve this, activating lockdown mode will limit some of the features provided by OpenAI through ChatGPT and other products. For example, you can use image creation and upload photos to ChatGPT, but you cannot retrieve images from the internet or display images within a response. The chatbot can't even download files to analyze, but you can manually upload documents if you want insights. Other features such as Deep Research and Agent Mode are completely disabled. “Lockdown mode does not change memory, file uploads, conversation sharing capabilities, or whether conversations can be used to improve models,” OpenAI adds. “Most of these settings can be configured separately by your workspace administrator.”

The company also points out that lockdown mode does not prevent immediate insertions from appearing in the ChatGPT content process. Instead, it is designed to prevent attackers from extracting sensitive data from your account by limiting the network requests that someone can exploit. Lockdown mode is available for all personal accounts, including those using ChatGPT through OpenAI's free tier. To activate, open ChatGPT's Settings menu and select Safety & Security. Under Advanced Security, tap Lockdown Mode and turn the toggle on. You can temporarily disable extra protection by selecting Manage from the status message that appears above the chat window and then selecting Turn off for this chat.

Separately, OpenAI is rolling out an active session manager that will allow users to see all devices or browsers used to access their accounts. Here, the company gives you the option to log out of individual sessions or all sessions at once. The latter may take up to 30 minutes to complete. “If you suspect unauthorized account activity, change your password if you use one, review your login method, and contact OpenAI Support,” the company added.