OpenAI Just Added a “Lockdown Mode” for ChatGPT

As AI tools become more powerful, one of the biggest security risks isn't the model itself.

It's what the model reads.

OpenAI has announced a new security feature called Lockdown Mode, designed to reduce the risk of prompt injection attacks — a growing threat where malicious instructions are hidden inside websites, documents, emails, or other content that an AI system accesses.

Think of it like a phishing attack, but aimed at the AI instead of the human.

A prompt injection can trick an AI into ignoring its original instructions, leaking sensitive information, or performing actions it shouldn't. As AI assistants gain access to company data, internal documents, and external tools, these attacks are becoming a serious concern for businesses and security teams.

That's where Lockdown Mode comes in.

When enabled, ChatGPT operates in a much more restricted environment. The feature disables live web browsing, blocks retrieval and display of images from the web, and turns off capabilities such as Deep Research and Agent Mode. Users can still generate images and interact with uploaded files, but the model's exposure to potentially malicious external content is significantly reduced.

OpenAI is careful not to oversell the protection.

The company notes that prompt injection attacks can still occur through cached web content or uploaded files, meaning Lockdown Mode is not a complete solution. Instead, it's designed to reduce the attack surface and lower the chances of sensitive information being exposed through manipulated content.

And that's an important distinction.

For the average ChatGPT user, these restrictions would likely feel unnecessary. But for organizations handling confidential business information, legal documents, financial records, research data, or intellectual property, the tradeoff may be worth it.

The launch reflects a broader shift happening across the AI industry.

For the past few years, the focus has been on making AI models smarter, faster, and more capable. Now, attention is increasingly turning toward security, governance, and trust. As AI systems gain access to more tools and data sources, protecting them from manipulation is becoming just as important as improving their intelligence.

Why it matters:

Prompt injection is quickly emerging as one of the biggest security challenges in the age of AI agents. OpenAI's Lockdown Mode signals that AI companies are beginning to treat these threats seriously, building dedicated safeguards for users who handle sensitive information. As businesses deploy AI deeper into critical workflows, security features like this may become as essential as the models themselves.