OpenAI has introduced a new Lockdown Mode for ChatGPT that limits internet connected features to help protect sensitive information from emerging AI security threats.
Quick Summary – TLDR:
- OpenAI is rolling out Lockdown Mode for ChatGPT users who handle sensitive data.
- The feature reduces the risk of data exfiltration caused by prompt injection attacks.
- Several capabilities including web browsing, Deep Research, Agent Mode, and file downloads are disabled when Lockdown Mode is enabled.
- OpenAI says the feature is designed for high security use cases and is not intended for every user.
What Happened?
OpenAI has announced the rollout of Lockdown Mode, a new security focused setting for ChatGPT that aims to reduce the risk of sensitive information being exposed through prompt injection attacks. The feature is becoming available to eligible users across Free, Go, Plus, Pro, and self serve ChatGPT Business accounts.
The new mode limits several tools that interact with the web or external services, creating a more restricted ChatGPT experience for users who prioritize security over convenience.
OpenAI is adding “LOCKDOWN MODE” to #ChatGPT.
— The Hacker News (@TheHackersNews) June 6, 2026
It won’t stop prompt injections.
It’s built to reduce what attackers want next: a way to leak your data out.
The mode limits tools that connect to the web or external services, including browsing, images, deep research, agent mode,… pic.twitter.com/n78xD4rl7h
OpenAI’s New Security Focus
As AI tools become more capable, they are also becoming more attractive targets for attackers. One growing concern is the rise of prompt injection attacks, where malicious instructions are hidden inside content processed by an AI system.
These hidden instructions can potentially influence how an AI behaves and, in some cases, create opportunities for sensitive information to be exposed. OpenAI describes prompt injection as a challenging and evolving problem that affects large language models across the industry.
To address this risk, the company developed Lockdown Mode as an additional layer of protection. According to OpenAI, the feature builds on existing safeguards such as sandboxing, URL based protections, monitoring systems, enforcement mechanisms, and enterprise security controls.
Rather than preventing prompt injections from appearing altogether, the goal is to make it significantly harder for sensitive information to leave ChatGPT through network connected features.
What Lockdown Mode Disables?
When Lockdown Mode is turned on, ChatGPT restricts several capabilities that could potentially be used to move information outside the platform.
The following features are disabled or heavily limited:
- Live web browsing, which is restricted to cached content only
- Image retrieval and image display in standard responses
- Deep Research
- Agent Mode
- Canvas networking
- File downloads for data analysis
Users can still upload files manually, generate images where supported, and continue using many core ChatGPT functions. However, the experience becomes more controlled and security focused.
OpenAI says some connected experiences, including shopping agent features and certain live connector functions, may also become unavailable while Lockdown Mode is active.
Not a Complete Solution
While Lockdown Mode strengthens security, OpenAI is clear that it is not a perfect defense.
The company says the feature is designed to substantially reduce the risk of prompt injection based data exfiltration, but it cannot guarantee that sensitive information can never be exposed. Risks may still exist through trusted applications, unforeseen feature combinations, or future attack techniques.
OpenAI also notes that prompt injections can still affect ChatGPT’s behavior. For example, a malicious instruction hidden inside an uploaded document could potentially influence a response or cause the model to provide inaccurate information.
Importantly, Lockdown Mode does not change how memory works, how files are uploaded, whether conversations can be shared, or whether chats may be used to improve models. Those settings remain separate.
Available Across Multiple ChatGPT Plans
Lockdown Mode is rolling out to eligible logged in users on:
- ChatGPT Free
- ChatGPT Go
- ChatGPT Plus
- ChatGPT Pro
- Self serve ChatGPT Business
Users can enable the feature through the Security section of ChatGPT settings. OpenAI also confirmed that Developer Mode and Lockdown Mode cannot be used simultaneously, meaning activating one automatically disables the other.
The rollout comes alongside another new security feature from OpenAI that allows users to review active ChatGPT sessions and remotely sign out of devices if suspicious account activity is detected.
Why This Matters?
The launch of Lockdown Mode highlights a growing reality in the AI industry. Security concerns are no longer theoretical. As AI systems gain access to more tools, data, and external services, protecting sensitive information becomes increasingly important.
Organizations in fields such as healthcare, finance, legal services, and enterprise operations often work with highly confidential information. For these users, sacrificing some convenience in exchange for stronger protections may be a worthwhile tradeoff.
SQ Magazine Takeaway
I think Lockdown Mode is one of OpenAI’s most practical security updates so far. AI companies have spent years adding more capabilities, but security risks grow every time an AI system gains access to the web, files, or external tools.
This feature acknowledges that reality. While it will not eliminate prompt injection threats entirely, it gives security conscious users a way to reduce risk when working with sensitive information. That is a smart step as AI becomes more deeply integrated into professional workflows.
