OpenAI releases 'Lockdown Mode' feature to protect sensitive data

OpenAI has announced a new feature that the company claims will provide additional protection against cyberattacks in the form of prompt injection. These cyberattacks occur through malicious chatbot instructions that are hidden within web pages and other content sources.

As reported by TechCrunch on Saturday (6/6), when ‘Lockdown Mode’ is activated within OpenAI’s services, it disables live web searching, meaning users can only access cached content or data temporarily stored on their devices. The company noted that even with ‘Lockdown Mode’ enabled, ChatGPT remains vulnerable to certain prompt injections. For instance, attacks can still emerge within cached web content or uploaded files, potentially affecting the behaviour or accuracy of responses.

‘Lockdown Mode is not intended for everyone,’ stated OpenAI. ‘It is designed for individuals and organisations that handle sensitive data and require stricter protection from the risks of data leaks related to rapid injection.’ Consequently, OpenAI stated that ‘Lockdown Mode’ is currently being rolled out to ChatGPT Business self-serve accounts, as well as eligible personal accounts.