OpenAI has launched Lockdown Mode, an optional security feature in ChatGPT designed to enhance user safety. This mode focuses on mitigating prompt injection attacks, where external entities exploit AI systems to manipulate responses or access sensitive data. Lockdown Mode is particularly aimed at high-risk users like executives and security teams and is accessible via ChatGPT Enterprise and related versions.
This feature restricts web browsing to cached content, preventing live requests, and disables specific capabilities, including image responses and data downloads. However, it does not entirely eliminate the risk of prompt injections. OpenAI has also implemented “Elevated Risk” labels on certain features to indicate potential security vulnerabilities. In Lockdown Mode, app interactions are controlled by admins, with elevated risk actions outlined for monitoring. Lockdown Mode will soon be available for consumers, while ongoing assessments will dictate the future of “Elevated Risk” labeling as security improves.
Source link
