OpenAI rolls out ChatGPT Lockdown mode to all users to block prompt injection data theft
OpenAI has expanded Lockdown mode to all ChatGPT plan tiers, including Free, Go, Plus, Pro, and Business users. The security feature blocks outbound network requests to prevent prompt injection attacks from stealing sensitive data, but disables live web browsing, Deep Research, and Agent mode.
OpenAI Rolls Out Lockdown Mode to All ChatGPT Users
OpenAI has expanded its Lockdown mode security feature to all ChatGPT users across Free, Go, Plus, Pro, and Business plans, according to ZDNET. The optional setting, which launched in February 2025 for Enterprise, Edu, Healthcare, and Teachers plans, aims to protect users from data theft through prompt injection attacks.
How Lockdown Mode Works
Lockdown mode blocks outbound network requests to the internet and external file services to prevent attackers from exfiltrating sensitive information through malicious prompts. The feature does not prevent prompt injection attacks themselves—attackers can still inject malicious commands that access cached web content or uploaded files.
What Gets Disabled
When enabled, Lockdown mode restricts the following capabilities:
- Live web browsing: ChatGPT can only access cached content, making search results potentially outdated or unavailable
- Web image retrieval: The model cannot display or retrieve images from the live web, though users can still upload images and request image generation
- Deep Research: The feature is completely disabled
- Agent mode: Unavailable in Lockdown mode
- Canvas networking: Code generated through Canvas cannot access network resources
- File downloads: ChatGPT cannot download files for analysis, but can still process uploaded files
Target Use Case
According to the report, Lockdown mode is designed for individuals and organizations handling sensitive or confidential information requiring additional security layers. The feature trades functionality for security by preventing live data connections that could be exploited by prompt injection attacks.
Availability and Activation
The feature is rolling out gradually across all ChatGPT accounts. Users can enable it through Settings > Security > Advanced Security > Lockdown Mode. A warning message explains the restrictions before activation.
What This Means
Lockdown mode represents a defensive approach to prompt injection rather than a solution. By cutting off live network access, OpenAI acknowledges that the underlying vulnerability—the ability to inject malicious instructions into prompts—remains unsolved. Organizations working with highly sensitive data now have a protection option, but at the cost of ChatGPT's web-connected capabilities. The tradeoff reveals the security challenges inherent in AI systems that blend user instructions with dynamic external data sources.
Related Articles
OpenAI restores chat sidebar in Mac app after user backlash over confusing redesign
OpenAI has updated its ChatGPT Mac app to restore direct access to chat conversations through a prominent sidebar toggle. The fix addresses user complaints following a July 10 redesign that replaced the native Mac client with an Electron-based app and buried the standard chat interface behind Work and Codex features.
OpenAI's GPT-5.6 Codex Bug Deletes User Files When Attempting to Override $HOME Environment Variable
OpenAI has identified a critical bug in GPT-5.6's Codex implementation that causes unexpected file deletions. According to Thibault Sottiaux, the issue occurs when the model attempts to override the $HOME environment variable to define a temporary directory but mistakenly deletes $HOME instead, particularly when full access mode is enabled without sandboxing protections.
OpenAI's GPT-5.6 Sol Adds Five Reasoning Effort Settings, Follows DeepSeep-R1 RLVR Training Method
OpenAI released GPT-5.6 Sol, a new reasoning model family that comes in three sizes with roughly five to six reasoning-effort settings each. The release follows the DeepSeek-R1 methodology of using reinforcement learning with verifiable rewards (RLVR), nearly two years after OpenAI's original o1 model popularized LLM-based reasoning.
Cline v4.0.9 Adds GPT-5.6 ChatGPT Models, Fixes Token Count Over-Reporting
Cline, the AI coding assistant VS Code extension, released v4.0.9 on July 16, 2024, adding support for GPT-5.6 ChatGPT subscription models. The update fixes a bug where token counts were over-reported from OpenAI-compatible providers due to improper handling of cumulative usage snapshots.
Comments
Loading...