OpenAI has introduced Lockdown Mode and Elevated Risk labels in ChatGPT to combat prompt injection attacks. These new protections aim to help users and organizations mitigate security risks as AI systems take on more complex tasks involving web browsing and connected applications.

Lockdown Mode is an optional, advanced security setting designed for highly security-conscious users like executives and security teams at prominent organizations. It tightly constrains how ChatGPT can interact with external systems to reduce the risk of data exfiltration through prompt injection attacks. For example, web browsing in Lockdown Mode is limited to cached content, preventing live network requests from leaving OpenAI's controlled network.

Elevated Risk labels will now appear consistently across ChatGPT, ChatGPT Atlas, and Codex for certain capabilities that may introduce additional security risks. These labels provide clear explanations of what changes, what risks may be introduced, and when that access is appropriate. The protections build on existing enterprise-grade security measures and are available for ChatGPT Enterprise, ChatGPT Edu, ChatGPT for Healthcare, and ChatGPT for Teachers. Admins can enable Lockdown Mode through Workspace Settings and maintain granular control over app permissions.