Introducing Lockdown Mode and Elevated Risk labels in ChatGPT

2026-02-12 · Source: OpenAI News · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy · Depth: Intermediate, short

Summary

OpenAI introduced two new security features on February 13, 2026, to mitigate prompt injection attacks in ChatGPT and related products. "Lockdown Mode" is an optional, advanced security setting for high-risk users, such as executives or security teams, available in ChatGPT Enterprise, Edu, Healthcare, and for Teachers. It restricts ChatGPT's interaction with external systems, for example, limiting web browsing to cached content to prevent data exfiltration. Additionally, "Elevated Risk" labels will now consistently appear for certain capabilities in ChatGPT, ChatGPT Atlas, and Codex that may introduce additional security risks, providing users with clearer guidance to make informed choices about using these features, especially with private data. These additions build upon existing protections like sandboxing and URL-based data exfiltration safeguards.

Key takeaway

For CTOs and security leaders evaluating AI deployment, you should consider implementing OpenAI's new Lockdown Mode for your most at-risk employees, such as executives or security teams, to significantly reduce prompt injection attack surfaces. This feature, available in enterprise-grade ChatGPT offerings, provides granular control over external interactions, enhancing data security. Additionally, educate your teams on the "Elevated Risk" labels to ensure informed decision-making when using connected AI capabilities with sensitive data.

Key insights

New OpenAI features enhance security against prompt injection by restricting external interactions and labeling risky capabilities.

Principles

• Constrain external interactions to reduce data exfiltration risk.
• Inform users about inherent risks in connected AI capabilities.

Method

Lockdown Mode deterministically disables or limits tools like live web browsing to cached content. "Elevated Risk" labels provide in-product guidance for features with unmitigated risks.

In practice

• Enable Lockdown Mode for high-security organizational roles.
• Review "Elevated Risk" labels before using connected AI features.

Topics

• Prompt Injection
• AI Security
• ChatGPT
• Data Exfiltration
• Cybersecurity

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Security Engineer, Security Engineer, MLOps Engineer

Related on AIssential

• OpenAI unveils Lockdown Mode to protect sensitive data from prompt injection attacks — OpenAI's Lockdown Mode offers enhanced, but not absolute, protection against prompt injection for sensitive data handlers by disabling certain features.
• Prompt-inject ChatGPT with any web page — Chatbots inherently struggle to differentiate instructions from data, making prompt injection an unfixable security vulnerability.
• OpenAI announces new advanced security for ChatGPT accounts, including a partnership with Yubico — Hardware security keys offer robust phishing protection for high-value AI accounts but require careful management.
• OpenAI patches ChatGPT for Mac after security incident — OpenAI patched its ChatGPT Mac app after a supply chain attack compromised employee devices, but user data remained safe.
• A Layered Security Framework Against Prompt Injection in RAG-Based Chatbots — A layered security framework significantly reduces prompt injection vulnerability in RAG chatbots by addressing attack vectors across the inference pipeline.
• OpenAI Help: Lockdown Mode — OpenAI's Lockdown Mode prevents data exfiltration from prompt injections by limiting outbound network requests, mitigating a key "Lethal Trifecta" risk.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by OpenAI News.