OpenAI Introduces Lockdown Mode for ChatGPT to Boost Security and Prevent Data Leaks
June 6, 2026
In Lockdown Mode, several capabilities are disabled or restricted, including live web browsing, Deep Research, Agent Mode, web image retrieval, external network access for code, and file downloads; image uploads and generation remain available where appropriate.
The mode reduces exposure to attackers exploiting network requests or connected tools, but does not stop prompt injections from appearing in content or affect conversation memory, file uploads, or model-improvement processes.
The article is authored by tech and global affairs commentator Syed Ziyauddin, who has a background with major media organizations.
Lockdown Mode reflects a broader industry trend toward giving users more granular control over AI access and capabilities, balancing usefulness with security as AI tools integrate with websites, documents, and business software.
The rollout aligns with ongoing efforts to provide granular security controls over AI interactions.
The rollout could accelerate enterprise AI adoption by addressing data leakage concerns, while organizations must implement robust data-handling policies, training, and monitoring to mitigate remaining risks.
OpenAI has rolled out Lockdown Mode for ChatGPT, a configurable security setting designed to curb data exfiltration and mitigate prompt-injection risks.
High- and medium-risk guidance helps admins decide which apps and actions to enable, emphasizing trust, visibility, and potential exfiltration paths.
In enterprises, conversation memory, file uploads, sharing settings, and model-improvement usage remain configurable, and OpenAI has added an active session manager to monitor devices and signed-in browsers.
Lockdown Mode began with enterprise and select users and is now broadly available across personal and business accounts.
The rollout started with enterprise users in February and now extends to personal and self-serve business accounts, per OpenAI’s updated communications.
Elevated Risk labels warn users before potentially risky actions, such as clicking external links or connecting to other services.
Summary based on 24 sources
Get a daily email with more Tech stories
Sources

TechCrunch • Jun 6, 2026
OpenAI unveils Lockdown Mode to protect sensitive data from prompt injection attacks
The Times Of India • Jun 6, 2026
ChatGPT gets a Lockdown Mode: What it does and who actually needs it
OpenAI Help Center
Lockdown Mode | OpenAI Help Center