The AI trust score helps users evaluate the maturity, safety, and operational health of AI assets. As an administrator, you can fine-tune the score range and weight values to ensure your AI trust scores reflect what matters most to your organization.
The following example image shows the AI Trust Score widget that is included on the asset pages of your AI-related Technology Assets.
For more details on the factors Collibra uses to calculate the score for each supported asset type, go to AI Trust Score: contributing factors.
If you're a business user and want more information on interpreting and improving a score, go to Interpreting and improving the AI trust score of an asset.
Prerequisites
- You've ensured that the AI Trust Score widget is included in the Page Editor layout for the following asset types:
- AI Use Case
- AI Agent
- AI Agent Version
- AI Base Model
- AI Model Version
- All subtypes of these asset types
By default, the widget is included in the asset page layout for all of these asset types.
Steps
- In
Settings, click AI Command Center. - Click the AI Trust Score tab.
The AI Trust Score settings page opens. - Enter the required information.
Setting Description Show AI Trust Score visualization Determines whether the AI trust score is shown on AI asset pages. Trust score range Defines the low (red), medium (orange), and high (green) trust score ranges used to visually classify AI model trustworthiness.
Trust score weights For each asset type, determine the weight for each contributing theme. Select an asset type tab to configure its theme weights. Higher weights give a theme greater influence on the overall score. The Total weighted score chart updates dynamically as you edit values. Every theme must have a weight value. Enter "0" to exclude a theme. Use Reset all to 1 to restore equal weighting.
The following table shows which themes are available per asset type. For descriptions of each theme, see Trust score themes below.
Theme Use case Base model Model version Agent Agent version Data assets 
Data integrity Documentation Lifecycle progress Linked technology assets Linked versions score Risk rating - Click Save.
The updated weights apply to subsequent trust score calculations. Scores for existing assets update automatically within seconds.
Trust score themes
Data assets
The criterion for this theme is met if a relation exists between the AI Use Case asset and at least one Data Asset.
Data integrity
The criteria for this theme are the following:
- Data lineage: A relation exists between the AI asset and at least one Data Asset, for example a Table asset or View asset.
- Data sensitivity: Related input data assets have a value (True or False) for the "Personally Identifiable Information" attribute type.
The contribution from this theme is derived as follows:
- If none of the criteria are met, the Data integrity score is 0.
- If one of the criteria is met, the Data integrity score is 60.
- If both of the criteria are met, the Data integrity score is 100.
Documentation
The criteria for this theme are the following:
- At least one user or user group is assigned the Owner resource role on the community or domain of the asset.
- On the asset page, there is a description in the Description field.
- A relation exists between the asset and a related asset type: for Use case assets, at least one Technology Asset; for all other asset types, at least one Use Case asset.
The Documentation score is derived as follows:
- If none of the criteria are met, the Documentation score is 0.
- If one of the criteria is met, the Documentation score is 40.
- If two of the criteria are met, the Documentation score is 60.
- If all of the criteria are met, the Documentation score is 100.
Lifecycle progress
The criterion for this theme is the advancement of the asset through its lifecycle, as determined in the Lifecycle tab in the global assignment of the asset type.
The contribution from this theme increases incrementally as the asset progresses through its lifecycle. For example, if 5 lifecycle statuses are configured for the asset type, the score increases by 20% for every successive stage reached.
Linked technology assets
The criterion for this theme is met if a relation exists between the AI Use Case asset and at least one Technology Asset.
The contribution from this theme is the average trust score of all Technology Assets related to the AI Use Case. For example, if the AI Use Case has a relation to 3 Technology Assets with AI trust scores of 20, 20, and 80, the contribution from this theme is (20 + 20 + 80) / 3 = 40.
Linked versions score
The contribution from this theme is the average trust score of all related:
- AI Model Version assets, as a contributing factor for AI Base Model assets.
- AI Agent Version assets, as a contributing factor for AI Agent assets.
For example, if an AI Agent asset has three related AI Agent Version assets with trust scores of 40, 60, and 80, the contribution from this theme is (40 + 60 + 80) / 3 = 60.
Note If an asset has no linked versions, this theme is excluded from the trust score and does not affect the score. As a result, this theme may appear on some AI Agent and AI Base Model assets but not others in the same instance. This is expected behavior.
Risk rating
The criterion for this theme is the answer to the "What is the overall risk rating of this AI use case?" question in the out-of-the-box Risks and Safeguards assessment.
The contribution from this theme is derived as follows:
- If the Risks and Safeguards assessment was not completed or no value was provided for the overall risk rating, the Risk rating score is 0.
- If the overall risk rating is set to "High" in the assessment, the Risk rating score is 10.
- If the overall risk rating is set to "Medium" in the assessment, the Risk rating score is 50.
- If the overall risk rating is set to "Low" in the assessment, the Risk rating score is 100.