Particle.news

OpenAI Adds Lockdown Mode to ChatGPT to Curb Data Exfiltration

The opt-in setting turns off web-facing tools to reduce the risk that hidden instructions in content can make the assistant leak sensitive data.

Overview

  • OpenAI rolled out Lockdown Mode on Saturday, June 6, 2026, as an optional, advanced security setting aimed at people and organizations that handle sensitive information.
  • When enabled, Lockdown Mode disables or restricts features that make outbound network calls, including live web browsing, image retrieval/display, Deep Research, Agent Mode, Canvas networking, and file downloads.
  • The feature is available across Free, Go, Plus, Pro and self-serve ChatGPT Business plans and can be turned on in ChatGPT’s Safety and Security settings.
  • OpenAI says Lockdown Mode lowers the chance of prompt-injection driven exfiltration but does not eliminate all risk and does not change conversation memory or file-upload behavior.
  • Security researchers call the change a blunt but practical defense and urge more surgical fixes such as scanning or sanitizing tool results before they reach the model so functionality can be preserved.