Filtering AI Services by Risk Type
Use the Artificial Intelligence (AI) Risk Type filter to sort AI services by High, Medium, and Low risk in the Skyhigh Cloud Registry. Sorting AI services by risk type allows you to take necessary actions to strengthen the defense against associated potential risks. SOCs can review and prioritize the AI services and outline remediations quickly, thereby enhancing the ability to secure their most sensitive data from high-risk AI services.
The AI services in Skyhigh CASB are associated with an AI risk score. The AI risk score of a service is measured based on the Large Language Model (LLM) used by the service and LLM risk attribute values (Jailbreak, Toxicity, Bias, and Malware). The LLM risk attribute values are determined based on the LLMs' risk weights.
Each service is categorized with an AI Risk score of 3, 6, 9, or NA, depending on the average value of the LLM risk attributes.
NOTE: The AI Risk score is calculated only if the AI service supports LLM.
The AI risk scores are determined by evaluating the various combinations of whether the service supports LLMs, the risk attribute values associated with LLMs, and the average values of these risk attributes. For a detailed representation of the AI risk score of a cloud service, refer to the table below:
| LLM Support | LLM Risk Attributes Value (Jailbreak, Toxicity, Bias, Malware) | Average value of the LLM Risk Attributes | AI Risk Score of a Service |
|---|---|---|---|
| Yes | High/Medium/Low | Low | 3 |
| Yes | High/Medium/Low | Medium | 6 |
| Yes | NPK (Not Publicly Known) | Medium | 6 |
| Yes | High/Medium/Low | High | 9 |
| No | NA | NA | NA |
Perform the steps below to filter AI services by risk type:
- On the Cloud Registry page, select Artificial Intelligence from the Service Category filter.
- Under Artificial Intelligence Risk Type, select the risk type. For example, we have selected Low Risk. The search results display only the low-risk AI services.
View AI Risk Score for Cloud Services
Enable the AI Risk option on the Edit Table Columns dialog (Actions > Edit Table Columns) to view the AI risk score for cloud services on the Cloud Registry page.
You can also enable AI related risk attributes on the Edit Table Columns dialog (Actions > Edit Table Columns) to view them on the Cloud Registry page.