Krux
February 21, 2026
ChatGPT's New Lockdown Mode Blocks Live Web Browsing
Published: February 21, 2026 at 3:44 AM
Updated: February 21, 2026 at 3:44 AM
100-word summary
OpenAI just dropped Lockdown Mode for ChatGPT, a high-security setting rolling out to Enterprise, Edu, Healthcare, and Teachers accounts, with consumer access coming soon. Designed for executives and security teams worried about data leaks, it tackles prompt-injection attacks by blocking live web requests and restricting external app access. Instead of browsing the open internet, ChatGPT can only pull cached content within OpenAI's network. Admins can customize which tools stay active per user. Paired with new Elevated Risk labels flagging dangerous features, this upgrade could reshape how organizations trust AI with sensitive workflows and confidential information.
What happened
OpenAI just dropped Lockdown Mode for ChatGPT, a high-security setting rolling out to Enterprise, Edu, Healthcare, and Teachers accounts, with consumer access coming soon. Designed for executives and security teams worried about data leaks, it tackles prompt-injection attacks by blocking live web requests and restricting external app access. Instead of browsing the open internet, ChatGPT can only pull cached content within OpenAI's network. Admins can customize which tools stay active per user.
Why it matters
Paired with new Elevated Risk labels flagging dangerous features, this upgrade could reshape how organizations trust AI with sensitive workflows and confidential information.