Reported by Gold 101.3 FM, UAE’s No.1 Malayalam radio station.
OpenAI has unveiled a new security feature for ChatGPT called Lockdown Mode, designed to help protect users and organisations from prompt injection attacks and reduce the risk of sensitive data being exposed.
The optional security setting is being rolled out across eligible ChatGPT accounts, including personal users, business subscribers, and enterprise customers.
According to OpenAI, Lockdown Mode is specifically designed to minimise the risk of data theft by restricting certain internet-connected features that could potentially be exploited by attackers.
What are prompt injection attacks?
Prompt injection attacks occur when malicious instructions are hidden within content processed by an AI system, such as websites, documents, files, or external data sources.
These hidden prompts can attempt to manipulate an AI model’s behaviour, potentially influencing responses or tricking the system into revealing sensitive information.
OpenAI says Lockdown Mode does not prevent such malicious instructions from appearing in content. Instead, it focuses on blocking one of the most critical stages of an attack — the unauthorised transfer of data outside the system.
How Lockdown Mode works
When Lockdown Mode is enabled, ChatGPT limits several features that rely on live internet access or external integrations.
Key restrictions include:
- Live web browsing is limited to cached content
- Deep Research becomes unavailable
- Agent Mode is disabled
- File downloads for data analysis are blocked
- External connector write actions are restricted
- Shopping-agent features are disabled
- Financial service integrations become unavailable
Users can still manually upload files, but some search results may be incomplete or less up-to-date due to the browsing restrictions.
Features that continue to work
OpenAI says Lockdown Mode does not affect several core ChatGPT functions, including:
- Memory features
- Conversation sharing
- Image generation
- Chat history settings
- AI model training preferences
The company also noted that Lockdown Mode does not impact network access within Codex.
However, users cannot activate Lockdown Mode and Developer Mode simultaneously. Enabling one feature will automatically disable the other.
How to enable Lockdown Mode
Eligible users can activate the feature directly within ChatGPT settings by following these steps:
- Open ChatGPT and go to Settings
- Select Security
- Navigate to Advanced Security
- Turn on Lockdown Mode
The feature is currently rolling out to eligible personal accounts, including Free, Go, Plus, and Pro plans, as well as self-serve ChatGPT Business subscriptions.
Growing focus on AI security
The launch highlights the increasing importance of cybersecurity and privacy protections as AI tools become more deeply integrated into personal and professional workflows.
As organisations use AI systems to process larger volumes of sensitive information, security features such as Lockdown Mode are expected to play a growing role in protecting users from emerging threats while maintaining access to powerful AI capabilities.
OpenAI says the new feature is part of its broader effort to strengthen user security and provide greater control over how information is accessed and handled within ChatGPT.
