OpenAI unveils Lockdown Mode to protect sensitive data from prompt injection attacks

OpenAI has launched a new security feature called Lockdown Mode to mitigate the risks associated with prompt injection attacks, where malicious instructions are embedded within web content or files. When enabled, this mode restricts certain functionalities, including live web browsing, external image retrieval, deep research, and agent mode, to minimize data exfiltration risks.

While OpenAI acknowledges that ChatGPT may still remain partially vulnerable to sophisticated injections within cached content or uploaded files, the feature is specifically targeted at organizations and users who handle sensitive information. The update is currently being rolled out to eligible personal accounts and self-serve ChatGPT Business customers.

(Source：TechCrunch)