OpenAI is expanding its security-focused Lockdown Mode to all ChatGPT users, including Free, Plus, and Business accounts, to prevent malicious actors from hijacking LLMs and stealing sensitive personal data.

Key Points

• Lockdown Mode blocks live outbound network requests to prevent unauthorized data exfiltration by malicious actors.
• The feature disables or restricts specific ChatGPT functions that rely on external web calls to enhance user privacy.
• OpenAI is rolling out the security setting to Free, Plus, Pro, Go, and self-serve Business account tiers.
• Users can enable the feature by navigating to the Advanced Security section within the ChatGPT settings menu.
• The tool was originally developed to counter prompt-injection attacks where hackers lace webpages with instructions to hijack AI models.

Why it Matters

This update provides individual users with enterprise-grade security tools to protect sensitive information from increasingly sophisticated AI-based cyberattacks. By democratizing access to these privacy controls, OpenAI aims to mitigate risks associated with LLM hijacking across its entire global user base.