OpenAI rolled out Lockdown Mode and Elevated Risk labels to counter prompt injection attacks by adding deterministic infrastructure controls that prevent data exfiltration regardless of prompt manipulation. The features restrict network interactions for high-risk users and flag network-related capabilities to help administrators and users manage exposure. #OpenAI #LockdownMode
Keypoints
- Lockdown Mode is an optional setting for high-risk users that enforces deterministic restrictions to block prompt injection exploits.
- The mode confines web browsing to cached content and prevents live network requests outside OpenAI’s controlled network.
- Capabilities without strong deterministic safety guarantees—like images, Deep Research, Agent Mode, network downloads, and Canvas-approved network code—are disabled.
- Workspace administrators enable Lockdown Mode via specialized roles and maintain granular app/action controls with visibility through Compliance API Logs Platform.
- Elevated Risk labels mark network-related features (such as Codex network access) to inform users of residual risks and are updated or removed as mitigations improve.
Read More: https://thecyberexpress.com/openai-new-lockdown-mode/