OpenAI has launched Lockdown Mode and Elevated Risk labels in ChatGPT to enhance security against prompt injection attacks and other threats. Lockdown Mode, an advanced setting, restricts ChatGPT’s interactions with external systems to minimize data exfiltration risks. For instance, web browsing capabilities are limited to cached content to prevent sensitive information from being leaked to attackers. Admins can activate Lockdown Mode via Workspace Settings and specify accessible apps for users. This feature is currently available in ChatGPT Enterprise, Edu, Healthcare, and for Teachers, with consumer options expected soon. Elevated Risk labels educate users about potential security risks when integrating AI features with external applications. These labels clarify the implications of enabling specific functionalities, helping developers make informed choices. OpenAI is committed to continuously enhancing security measures and will update risk labels as they improve safeguards over time. This proactive approach aims to protect users while promoting transparency in AI usage.
Source link
