Back Gigazine

OpenAI announces lockdown mode to protect users' sensitive information from prompt ...

Gigazine • June 7, 2026

OpenAI has announced ' Lockdown Mode' to protect users' sensitive information from

Lockdown mode is available for all account types and workspaces, but it is being rolled out gradually, so it may not yet be displayed in some users' settings screens. Lockdown mode can be found in ChatGPT settings under 'Security' > 'Advanced Security'. Note that Lockdown mode and Developer mode cannot be enabled simultaneously. Also, when Lockdown mode is enabled, a status message will be displayed above the Composer. To turn it off for a specific chat only, select 'Manage' from the status message and then select 'Turn off for this chat'. Alternatively, you can open 'More Options', select 'Lockdown', and then select 'Disable'.

In essence, a prompt injection attack is a type of cyberattack against AI that involves manipulating the AI to leak confidential information or spread misinformation by disguising malicious input as a legitimate prompt. OpenAI is continuously strengthening its multi-layered security and safety systems to protect users' sensitive information from prompt injection attacks, and lockdown mode is one such system based on protection at the AI model, product, and system levels. Specifically, this includes enterprise controls such as sandboxing,

In lockdown mode, features such as memory usage, file uploads, conversation sharing, and whether conversation content is used to improve the AI model remain unchanged. Many of these settings can be configured individually by the workspace administrator. According to OpenAI, enabling lockdown mode does not change the settings regarding whether or not conversation content is used to train the AI model.

Jun 07, 2026 13:00:00 in AI , Security , Posted by logu_ii

The full story

This article is one source in a clustered incident — the cluster page carries the summary, timeline and every other outlet covering it.

OpenAI Launches Lockdown Mode to Combat Prompt Injection Attacks OpenAI has introduced Lockdown Mode, an advanced security feature for ChatGPT, aimed at mitigating prompt injection attacks. This mode limits the AI's ability to connect to the web or external services, significantly reducing the risk of data exfiltration. Prompt injection attacks involve embedding malicious instructions within content that AI systems process, potentially leading to unauthorized data access. Lockdown Mode disables features like live web browsing, Deep Research, and Agent Mode, while still allowing users to upload files. Initially launched for enterprise users in February 2026, it is now available to all personal accounts, including Free, Plus, Pro, and self-serve Business accounts. OpenAI emphasizes that while Lockdown Mode enhances security, it does not completely eliminate the risk of prompt injections. Users are encouraged to enable this feature if they handle sensitive information. 39 sources 55

Extracted Entities

Attack Types (1)

Prompt Injection Attack