OpenAI has launched a new feature for its AI chatbot ChatGPT, known as Lockdown Mode, aimed at enhancing protection against prompt injection attacks. These attacks involve malicious instructions hidden within web pages and other content that can trick the AI into executing unauthorized actions. Lockdown Mode significantly reduces the risk of data leakage by restricting outgoing network requests and disabling certain functionalities. For instance, the feature turns off real-time web browsing, allowing access only to cached content, and limits the retrieval of images and advanced search capabilities.
While OpenAI emphasizes that this mode is designed to mitigate the risks associated with prompt injections, it does not guarantee complete data protection. Risks may still arise from enabled applications, and cached web content or uploaded files could still influence the AI's behavior and response accuracy. The company clarified that Lockdown Mode is specifically tailored for individuals and organizations dealing with sensitive information who require stricter safeguards against potential data leaks.
Currently, the Lockdown Mode is available to ChatGPT Business accounts and select personal accounts, including Free, Go, Plus, and Pro. This move strategically positions OpenAI to address growing concerns about data security and privacy in the AI landscape, potentially giving it an edge over competitors in the market.
Informational material. 18+.
