OpenAI has introduced two new security features for ChatGPT users: Lockdown Mode and “elevated risk” warnings, aimed at strengthening protections for high-risk users and improving transparency around potentially sensitive tools.
Lockdown Mode is an optional advanced security feature designed primarily for executives, security professionals and enterprise teams who may face heightened cybersecurity threats. It is not intended for general users.
When activated, Lockdown Mode restricts how ChatGPT interacts with external systems. Certain tools that could be vulnerable to prompt injection attacks are deterministically disabled. Web browsing is limited to cached content, meaning no live network requests leave OpenAI’s secure environment. Features that cannot guarantee strong data protection may be turned off entirely.
Enterprise administrators can enable Lockdown Mode by assigning a dedicated role within workspace settings. They can also customise which applications and specific actions remain available while the feature is active. Additional compliance logging tools provide detailed insights into app usage, shared data and connected systems.
The company said Lockdown Mode is available for enterprise and sector-specific plans, with consumer access expected in the coming months.
OpenAI also introduced “elevated risk” labels for certain features in ChatGPT, ChatGPT Atlas and Codex. These warnings are designed to clarify when a tool may involve emerging or higher-risk capabilities, such as accessing external networks through coding utilities.
The company said the labels aim to help users make informed decisions, particularly when handling sensitive data. As security protections evolve, some features may eventually lose the elevated risk designation, while new capabilities could be added to the category.
OpenAI said the updates reflect its broader effort to strengthen safeguards as AI tools become more deeply integrated into workplace and enterprise environments.
