Hackers exploit prompt injection attacks to infiltrate AI tools, threatening the confidentiality of user data. In response, OpenAI has launched ChatGPT’s Lockdown Mode, enhancing security by restricting interaction with external systems, thus minimizing the risk of data exfiltration. This optional feature is particularly aimed at security-conscious users like executives and healthcare professionals, available in plans such as ChatGPT Enterprise and Edu. Lockdown Mode limits browsing capabilities and disables risky features unless data safety is confirmed. To further safeguard users, OpenAI has introduced Elevated Risk labels that alert users to potentially dangerous tools or actions in applications like ChatGPT and Codex. These labels serve as a preliminary warning, with plans for more robust security enhancements in the future. By implementing these features, OpenAI aims to protect sensitive information and mitigate the growing threat of cyberattacks linked to AI technologies.
Source link
