OpenAI launches a new feature, Lockdown Mode, designed to provide extra security for high-risk profiles.
OpenAI is introducing a new security feature for ChatGPT called ‘Lockdown’ mode, which offers additional protection for users with an elevated risk profile. This optional setting limits the model’s interaction with external systems: certain tools are disabled, and web browsing is conducted exclusively via cached content.
With this, OpenAI aims to provide protection against prompt injection attacks, where attackers try to manipulate the model into revealing sensitive information. The feature builds on existing security within ChatGPT enterprise subscriptions and will be rolled out to consumers in the coming months.
Lockdown
The new Lockdown Mode is an optional security feature in ChatGPT intended for users at high risk of targeted attacks, such as executives or security teams. The mode reduces the chance of data exfiltration by strictly limiting interaction with external systems: certain tools are disabled, web access occurs only via a cache without active network requests, and features whose security cannot be guaranteed are removed entirely.
The feature builds on existing enterprise security and can be enabled by administrators within ChatGPT Enterprise and similar business subscriptions. In the coming months, the feature will also become available to consumers.