Lockdown Mode and Elevated Risk Labels in ChatGPT (openai.com)

OpenAI has introduced two significant security measures in ChatGPT to combat the rising threat of prompt injection attacks, where malicious third parties manipulate AI systems to extract sensitive information. The first measure, Lockdown Mode, is an advanced security option tailored for high-risk users, such as executives, allowing for tighter control over how ChatGPT interacts with external systems. In this mode, features like web browsing are restricted to cached content, effectively preventing potential data exfiltration through live network requests. This mode is designed to enhance existing enterprise-grade data security offered through ChatGPT's business plans. The second measure involves implementing "Elevated Risk" labels for specific functionalities across ChatGPT, ChatGPT Atlas, and Codex. This labeling standardizes the communication of risk associated with certain features, enabling users to make informed decisions about engaging with capabilities that may introduce additional vulnerabilities. For instance, the coding assistant Codex can access the web for documentation, but it now carries an “Elevated Risk” label to alert developers to the associated risks. As OpenAI continues to enhance its safety protocols, the plan is to phase out these labels once risks are sufficiently mitigated, reflecting the ongoing commitment to user security in an increasingly complex AI landscape.