TEMPO.CO, Jakarta - The tech giant OpenAI is rolling out a new "Lockdown Mode," intended to provide additional protection against prompt injection attacks—the method of concealing dangerous instructions for chatbots on the web or other content sources.
Through this feature, OpenAI disables several capabilities of ChatGPT, from live web browsing and fetching images from the internet to deep research, and agent mode.
Users can only access cached content and create images through regular prompts.
OpenAI acknowledges that Lockdown Mode does not entirely eliminate the risk of prompt injection. The ChatGPT service still faces potential impact from such attacks, for example, through web cache content or user-uploaded documents.
"But the goal is to reduce the likelihood that sensitive data gets shared in the process," says OpenAI, as quoted from Tech Crunch, Saturday, June 6, 2026.
The company explains that the launch of Lockdown Mode is starting for ChatGPT Business self-serve accounts and eligible personal accounts. This feature is designed for specific users, especially those handling sensitive information and requiring additional protection against data leakage risks.
"It is designed for people and organizations that handle sensitive data and want stricter protection from data exfiltration risks related to prompt injection," reads the statement.
"Lockdown Mode is not intended for everyone," it added.
Read: Indonesia Scraps OpenAI from Digital VAT Collector List