OpenAI launches Lockdown Mode, offering ChatGPT users extra protection from hidden prompts

OpenAI has unveiled a new security-focused feature for ChatGPT aimed at reducing the risk of prompt injection attacks, a growing concern as AI assistants become increasingly connected to external content and tools.

Called Lockdown Mode, the feature is designed for users who handle sensitive information and want stronger protections against attempts to manipulate AI systems through hidden instructions embedded in websites, documents or other sources. While OpenAI acknowledges that the setting does not eliminate the threat entirely, it says the new mode significantly limits the pathways through which such attacks can occur.

ChatGPT Lockdown Mode: What is it?

Prompt injection attacks occur when malicious instructions are concealed within content that an AI model reads or processes. These hidden commands can potentially influence the chatbot's behaviour, leading it to produce inaccurate responses or, in some cases, expose information it should not share.

To reduce those risks, Lockdown Mode disables several of ChatGPT’s most powerful connected features. Users who enable the setting will no longer be able to access live web browsing, meaning ChatGPT can only rely on cached content rather than retrieving information directly from the internet. The mode also blocks the retrieval and display of images from the web, although image generation tools remain available.

More from Tech

In addition, advanced capabilities such as Deep Research and Agent Mode are switched off. These tools are designed to gather information, perform tasks and interact with external resources on behalf of users, making them particularly useful but also increasing the potential attack surface for malicious prompts.

OpenAI says the feature is intended for a relatively narrow audience rather than everyday users. According to the company, Lockdown Mode is aimed at individuals and organisations that regularly work with confidential or sensitive information and want tighter controls over how ChatGPT interacts with external content.

The company has also been careful to stress that the feature is not a complete defence against prompt injection attacks. Hidden instructions could still exist within cached web pages or uploaded files, potentially affecting how the chatbot responds. However, OpenAI argues that reducing direct exposure to live external sources lowers the chances of sensitive information being leaked through manipulation attempts.

The rollout is currently under way for self-service ChatGPT Business customers as well as eligible personal account holders.

The launch highlights a broader shift within the AI industry, where security concerns are increasingly taking centre stage as companies race to build more capable digital assistants. While AI models are becoming better at carrying out complex tasks, they are also being given greater access to tools, websites and third-party services, making robust safeguards increasingly important.

OpenAI plans to overhaul ChatGPT app into 'super app'

The introduction of Lockdown Mode comes as reports suggest OpenAI is pursuing a much larger transformation of ChatGPT.

According to the Financial Times, the company wants to evolve ChatGPT from a conversational assistant into a comprehensive AI-powered “super app” capable of helping users manage both professional and personal tasks.

OpenAI executive Thibault Sottiaux reportedly described the long-term vision as a personal AI agent that can support users across a wide range of activities, from workplace productivity and software development to everyday life administration.

A central part of that strategy is Codex, OpenAI’s coding platform, which has already amassed more than five million weekly active users. The company believes task-completing AI agents, capable of handling activities such as organising schedules, managing workflows and making bookings, could eventually become more valuable than traditional chatbots focused primarily on conversation.

To support those ambitions, OpenAI is reportedly redesigning both the web and mobile versions of ChatGPT. Future updates are expected to place greater emphasis on coding tools, image creation features and integrations with partner applications. The company is also said to be restructuring internal product teams while placing increased focus on enterprise customers.