ChatGPT Enhances Security with New Feature to Counter Prompt Injection Attacks
OpenAI Introduces Enhanced Security Measures to Combat Prompt Injection Attacks in ChatGPT
OpenAI has rolled out a new security feature, Lockdown Mode, and introduced Elevated Risk labels in ChatGPT to help mitigate the risks associated with prompt injection attacks and other advanced security threats. These enhancements are particularly relevant when using features that interact with external systems.
Lockdown Mode
Lockdown Mode is an optional, advanced security setting designed for highly security-conscious users who require protection against sophisticated threats. This feature constrains how ChatGPT interacts with external systems by disabling certain tools and capabilities that an attacker could potentially exploit through user conversations or connected apps. For instance, it limits the ability of ChatGPT to interact with external systems, thereby reducing the risk of data exfiltration via prompt injection attacks.
To enable Lockdown Mode, administrators can create a dedicated role through Workspace Settings, which adds security restrictions on top of existing administrative controls. This allows admins to choose which apps and actions are available to users operating in Lockdown Mode. Currently, Lockdown Mode is available for ChatGPT Enterprise, ChatGPT Edu, ChatGPT for Healthcare, and ChatGPT for Teachers, with plans to expand its availability to consumer users in the future.
Elevated Risk Labels
In addition to Lockdown Mode, OpenAI has introduced Elevated Risk labels, which provide in-product guidance for features that may introduce additional security risks when connecting AI products with apps and the web. These labels help users understand the changes and risks involved, enabling them to make informed decisions. Elevated Risk labels are applied to features across ChatGPT, ChatGPT Atlas, and Codex, and explain what a feature does, what changes when it is enabled, what risks may be introduced, and when its use is appropriate.
For example, in Codex, granting network access allows the system to take actions on the web. When enabled, an Elevated Risk label highlights the security implications of this access, enabling developers to better understand the trade-offs involved.
According to OpenAI, “We continue to invest in strengthening our safety and security safeguards, especially for novel, emerging, or growing risks. As we strengthen the safeguards for these features, we will remove the ‘Elevated Risk’ label once we determine that security advances have sufficiently mitigated those risks for general use. We will also continue to update which features carry this label over time to best communicate risk to users.”
These enhanced security measures demonstrate OpenAI’s commitment to protecting users from advanced security threats and providing them with the necessary tools to make informed decisions about their AI interactions.
About Author
English