Saved View to Identify AI Services with High-Risk LLM Attributes

Last updated
Save as PDF

Limited Availability: To access Skyhigh Recommended Saved Views for LLM Risk Attributes, contact Skyhigh Support.

Skyhigh CASB provides a recommended saved view for real-time visibility into LLM-based AI services, including Jailbreak, Malware, National Institute of Standards and Technology (NIST), and Open Worldwide Application Security Project (OWASP) risk attributes. Skyhigh Cloud Registry offers a comprehensive list of over 550 AI services. However, by utilizing a predefined saved view in Skyhigh CASB, you can enhance visibility and effectively track the specific AI services used in your organization.

Use these recommended saved views to monitor potential risks associated with LLM-based AI services, including Jailbreak, Malware, NIST, and OWASP risk attributes, identify unusual trends or security threats, and take necessary actions, such as blocking specific AI services, to ensure compliance with security standards.

You can access the Skyhigh-recommended saved view on the Services and Users pages. The following are the Skyhigh recommended LLM risk attribute saved views:

High Risk Jailbreak AI Services Usage. Use this predefined view to identify AI services associated with the High-Risk Jailbreak LLM risk attribute.
High Risk Malware AI Services Usage. Use this predefined view to identify AI services associated with the High-Risk Malware LLM risk attribute.
High Risk NIST Services Usage. Use this predefined view to identify AI services associated with the High-Risk NIST LLM risk attribute.
High Risk OWASP Services Usage. Use this predefined view to identify AI services associated with the High-Risk OWASP LLM risk attribute.

The saved views enable you to track usage metrics for the LLM's Jailbreak, Malware, NIST, and OWASP risk attributes. When you select one of these saved views, the corresponding page, either Users or Services, displays the filtered results for that view. For instance, if you select the High Risk Jailbreak AI Services Usage saved view on the Users page, the table shows high-risk users who have attempted jailbreak prompts or are linked to jailbreak activities. Similarly, on the Services page, the Services table lists high-risk services that have detected jailbreak prompts or interactions and marked the output as risky or manipulated.

Skyhigh CASB provides four default AI dashboard cards for LLM's Jailbreak, Malware, NIST, and OWASP risk attributes. For details, see Monitor AI Service Risks Using the AI Dashboard.

For details about the LLM risk attributes, see LLM Risk Attributes for Skyhigh Cloud Registry.