What is Lockdown Mode in ChatGPT?

Lockdown Mode is an optional security feature in ChatGPT designed for users with high privacy requirements, which strictly limits the AI's interaction with external systems.

How does ChatGPT address prompt injection attacks?

ChatGPT is introducing new security features, including Lockdown Mode and clearer risk labeling, to defend against prompt injection attacks that attempt to trick the AI into malicious actions.

What does the 'Elevated Risk' label mean on ChatGPT features?

The 'Elevated Risk' label indicates that a feature, such as those granting AI network access, poses an increased security risk and requires user awareness.

Home / Technology / New AI Shields ChatGPT from Malicious Prompts

New AI Shields ChatGPT from Malicious Prompts

18 Feb

•

Summary

ChatGPT introduces Lockdown Mode for enhanced user privacy.
New security labels warn users of elevated risk features.
Updates target prompt injection attacks tricking AI systems.

New AI Shields ChatGPT from Malicious Prompts

OpenAI is enhancing ChatGPT's security with two new features designed to protect against sophisticated AI threats. These updates directly address prompt injection attacks, where malicious prompts attempt to manipulate AI into performing harmful actions or revealing sensitive data.

One significant addition is Lockdown Mode, an optional security setting for users prioritizing privacy. This mode imposes strict limits on how ChatGPT interacts with external systems. Features enabling direct network calls are disabled, and web browsing relies on cached content.