Find the insights and best practices about our product.
AI System Details Overview

Introduction

The AI System Details page provides a comprehensive view of a single AI System within the Cranium platform. After selecting an AI System from My AI Systems, users are brought to this page to explore and manage the system's context, governance, artifacts, and risks.

The page is organized into five tabs: Overview, Published AI Cards, Vulnerabilities, Governance and Settings. Each tab surfaces a specific layer of information about the AI System. Together, these tabs help organizations track compliance, enrich documentation, and monitor security signals with precision. This article provides an overview of each tab and the type of information and actions available within them.

Tabs Overview

AI System Overview

The AI System Overview tab offers a quick summary of an AI System's type, description, and owner, aiding system management and collaboration. It displays core system details and provides artifact overview cards for all attached Bills of Materials, Compliance Frameworks, Vulnerability Reports, and Documents.

Published AI Cards

The Published AI Cards tab displays all AI Cards that have been published for this AI System, showing the AI Card name, AI System name, and publication date for each card. This tab also indicates when published AI Cards are out of sync with their source BOM and provides options to update individual cards or all out-of-sync cards in bulk.

Vulnerabilities

The Vulnerabilities tab shows the model and technology vulnerabilities for every Bill of Materials in the AI System together, with severity totals and filters.

Governance Information

The AI System Governance tab allows you to manage governance and compliance for the AI system, consisting of four subsections: System Information, Values, Data, and Governance Contact. This tab centralizes critical governance data to provide a structured overview of compliance and oversight activities related to the AI system.

AI System Settings

In the AI System Settings tab, users can update essential details like the AI System Name, Description, and Owner. These fields ensure proper identification and attribution, streamlining management workflows and maintaining accuracy.

AI System Overview

The Overview tab is a vital introduction to the AI System's core components and context. It details the system's name, owner, and description, as well as the latest publication date and time.

This section provides key system details and displays summary cards for all artifacts attached to the AI System. Users can export system information or initiate new AI Card publications directly from this view. The artifact cards allow you to review, perform actions, or add new artifacts to enhance documentation and ensure compliance with evolving standards.

The Overview tab is crucial for understanding an AI System's state and context. It streamlines management, promotes collaboration, and provides compliance and security insights.

AI System Details

The AI System Overview provides a comprehensive look at the system's core details. AI Systems in the Cranium platform serve as containers for the artifacts that represent their functionality, compliance, and security. Key information includes the system's name, owner, description, and the date and time of its most recent publication.

  • Owner: The system owner ensures compliance with governance, security, and regulatory standards in Cranium, maintaining documentation, monitoring risks, and acting as the primary contact.
  • Description: A concise AI system description should include its purpose, functionality, scope, limitations, and intended audience to ensure clarity and alignment with stakeholder expectations.

Artifact Overview

The artifact overview features identical artifact summary cards to those in the My AI Systems List, including:

  • Bills of Materials: Shows the most recent scan status and BOM details
  • Compliance Frameworks: Displays compliance framework status and scores
  • Vulnerability Reports: Shows Vulnerability Reports generated before new reports were retired. New Vulnerability Reports can no longer be created.
  • Documents: Provides access to supporting documentation

Users can review, perform actions, or add new artifacts to enhance documentation and ensure compliance with evolving standards.

Key Actions & Functionality

Add Artifacts

The Add Artifacts button allows you to select Compliance Frameworks, Bills of Materials, and Documentation to attach to the AI System.

Remove Artifacts

The 3-dot menu for each artifact includes the option to remove the selected artifact from the AI System. For Bills of Materials and Documentation, the artifact is removed from the system but not permanently deleted. You can still find the removed BOM or document in the Bill of Materials List and Document Library, respectively. For Compliance Frameworks and existing Vulnerability Reports, the artifact is removed and permanently deleted in the platform.

View Artifacts

The 3-dot menus for Bills of Materials and Compliance Frameworks include the option to "View Details" for the selected artifact. For Bills of Materials, this option will display the BOM inventory page. For Compliance Frameworks, this option opens the framework editor, where you can review, edit, and score your responses.

Download Artifacts

The 3-dot menus for existing Vulnerability Reports and Documentation include the option to download the selected artifact from the AI System. Existing Vulnerability Reports download in PDF format. Compliance Framework PDF exports include an executive summary page showing section scores without detailed questions and answers. The summary appears first in the PDF, followed by the complete framework details.

Export to PDF

Download a PDF summary of the AI System details, governance information, and models.

Publish New AI Card

Initiate AI Card publication to share the system with external organization.

Published AI Cards

The Published AI Cards tab displays all AI Cards that have been published for this AI System. Each entry shows the AI Card name, the AI System name, and the date the card was published. This tab provides a complete history of transparency documentation shared externally for this system, along with controls to keep published AI Cards aligned with the current state of their source BOM.

Out of Sync Status

An AI Card becomes out of sync when its associated BOM has been edited or rescanned since the card was published. The card displays an Out of Sync status with an Update button. When all published AI Cards on this tab are in sync with their source BOMs, the status updates and review controls do not appear.

Update an Individual AI Card

Click the Update button on an out-of-sync AI Card to open the Update [Card Name] to Latest Version modal. The modal lists each Bill of Materials contained in the card, with separate Models and Technologies sections under each. Changes are color-coded according to the legend at the top of the modal:

  • Net New Item (blue) indicates a component added to the BOM since publication.
  • Item Removed (red) indicates a component removed from the BOM.
  • Attribute Changed (purple) indicates a component whose attributes have changed.

Review the changes and click Confirm Update to synchronize the AI Card with the current BOM, or click Cancel to close the modal without updating.

Update All Out of Sync Cards

When two or more AI Cards on the tab are out of sync, an Update all Out of Sync Cards button appears at the top of the tab with a count of out-of-sync cards in parentheses. Click the button to open a confirmation prompt and then click Confirm to update all out-of-sync cards at once. This bulk action skips the change review modal. To review changes before updating, use the Update button on individual cards instead.

Vulnerabilities

The Vulnerabilities tab shows every vulnerability found in the AI System's Bills of Materials in one view. Use it to see where risk sits across the system and decide what to address first. The Models and Technologies sub-tabs have the same columns, filters and severity totals as the main Vulnerabilities page, limited to this AI System's Bills of Materials.

All Bills of Materials in the AI System appear together. To focus on one, filter by Bill of Materials. Other Vulnerabilities are not shown on this tab. They remain on the Bill of Materials details page.

Reviewing Model Vulnerabilities

The Models sub-tab lists each model in the AI System's Bills of Materials that has vulnerabilities. Each row shows:

  • BOM: The Bill of Materials that contains the model.
  • AI System: The AI System the Bill of Materials belongs to.
  • Model Name: The model's name. Click it to open the model's details in the Arena.
  • Detected: The date and time the vulnerabilities were detected.
  • Vulnerability Likelihood: The model's vulnerability likelihood score, shown as a percentage.
  • Vulnerable Attack Categories: A colored tag for the highest severity found, with the number of attack categories at that severity in parentheses. Categories at lower severities appear as links beside the tag. For example, MEDIUM (7) with 1 Low means seven medium-severity categories and one low-severity category.
  • Actions: The Resolve Vulnerabilities button.

Reviewing Technology Vulnerabilities

The Technologies sub-tab lists each vulnerable technology in the AI System's Bills of Materials. Each row shows:

  • BOM: The Bill of Materials that contains the technology.
  • Project Name: The Azure DevOps project or GitLab group the technology belongs to.
  • Repository Name: The repository the technology was found in.
  • Source: The package that contains the vulnerability.
  • Language: The package's language and ecosystem, shown as Language (Ecosystem), for example Python (PyPI). The ecosystem is PyPI, npm, Maven, NuGet, Go or crates.io.
  • Detected: The date and time the vulnerability was detected.
  • Description: A description of the package.
  • CVSS: The vulnerability's CVSS score.
  • Severity: The severity classification.
  • Identifier: The security identifier, such as a CVE, PYSEC or GHSA number. It links to the entry in the OSV database.
  • Version: The affected version. Hover over the info icon to see the version in which the vulnerability was fixed.
  • Dependency Type: Direct for a dependency the project declares itself, or Transitive for a dependency pulled in by one of those direct dependencies.
  • AI: Whether the vulnerability is on an AI-based technology. A purple AI tag marks an AI vulnerability, No marks a technology that is not AI-based, and a dash marks a technology that was not evaluated.
  • Actions: The Ignore and Resolve buttons.

A package name that exists in two ecosystems appears once for each ecosystem, so their advisories are not mixed.

Finding AI Vulnerabilities

A technology is AI-based when it matches Cranium's AI package list, and vulnerabilities on AI-based technologies are AI vulnerabilities. Technologies in Python, JavaScript and TypeScript, Java, C#, Go and Rust are evaluated. Technologies in other languages are not evaluated and always show a dash.

To list only AI vulnerabilities, open Filters, select Show AI vulnerabilities only under AI vulnerabilities, and click Apply Filters. The filter leaves out technologies that were not evaluated. The severity totals change to count only the AI vulnerabilities.

A Bill of Materials' labels are set when it is saved or scanned. They update on its next scan or edit, not when Cranium's AI package list changes.

Severity Totals

Severity totals show how this AI System's vulnerabilities are spread across severity levels. The Models sub-tab shows MODELS W/ VULNERABILITIES, and the Technologies sub-tab shows TECHNOLOGY VULNERABILITIES. Each total has a breakdown by severity: Critical, High, Medium, Low and Unknown.

The totals count only this AI System and follow the table's filters and search. Use the Show Severity Statistics toggle to turn them on or off.

Narrowing the View

Use Search and Filters to focus on the vulnerabilities that matter most. On the Technologies sub-tab, Filters includes:

  • Severity Level: Show only the selected severity levels.
  • Bill of Materials: Show only the selected Bills of Materials in this AI System.
  • Dependency Type: Show only Direct or Transitive dependencies.
  • Dependency groups: Show only packages declared in the selected groups of their manifest. A package whose manifest does not record a group shows as Unknown. The Prod only and Dev only buttons apply the matching group filter in one click.
  • AI vulnerabilities: Show AI vulnerabilities only.

On the Models sub-tab, filter by model name, severity or Bill of Materials. The Bill of Materials, model name and dependency group lists show only this AI System's entries.

Use Customize Table to reorder or hide columns. Columns hidden on this tab are also hidden on the main Vulnerabilities page, and the reverse. Filters set on the main Vulnerabilities page are cleared when you open this tab, and filters set here are cleared when you open that page.

Exporting Vulnerabilities

Export this AI System's vulnerabilities as a CSV file to share them or work with them outside the platform. The export covers only this AI System. It works the same way as the export on the main Vulnerabilities page.

  1. Open the Models or Technologies sub-tab.
  2. Click Export Vulnerabilities at the top right of the sub-tab, next to the Show Severity Statistics toggle. The Export Vulnerabilities dialog opens, titled with your organization's name.
  3. Under Configure Vulnerabilities Export Data, choose what to export:
    • Export all data: All of the AI System's vulnerabilities on that sub-tab. This option is selected by default.
    • Export filtered data: Only the vulnerabilities that match the filters and search applied on the sub-tab.
  4. Click Export, or click Cancel to close the dialog without exporting.

A confirmation message appears. The file is available in the Document Library when it is complete.

On the Models sub-tab, Export filtered data does not apply the severity filter. The export includes models of every severity that match the other filters.

What the Export Contains

Exporting from the Models sub-tab produces a CSV of model vulnerabilities. The CSV includes a Remediation column with category-level mitigation guidance for each finding.

Exporting from the Technologies sub-tab produces a CSV of technology vulnerabilities, including the direct or transitive classification. The RepositoryName column contains only the repository name. The CSV also has these columns:

  • Language: The package's language.
  • Ecosystem: The package's ecosystem: PyPI, npm, Maven, NuGet, Go or crates.io.
  • EcosystemInferred: TRUE when Cranium could not read the ecosystem from the project and used a best guess. FALSE when the project reported the ecosystem or it came from a known language. Read this column only when Ecosystem has a value.
  • ProjectName: The Azure DevOps project or GitLab group the technology belongs to. Blank where the source control system has no equivalent.
  • AISystemName: Each AI System that contains the technology, separated by semicolons. Deleted AI Systems are not listed. Blank where the source control system has no equivalent.
  • RepositoryOrganization: The GitHub owner, Bitbucket workspace, or Azure DevOps organization that owns the repository. Blank where the source control system has no equivalent.

In both CSV exports, a text cell that begins with =, +, -, @, a tab, or a carriage return starts with an apostrophe. The apostrophe stops spreadsheet applications from treating the cell as a formula. For example, @angular/core appears as '@angular/core. Spreadsheet applications hide the apostrophe, but scripts that read the raw file see it.

When No Vulnerabilities Appear

When the AI System has no Bills of Materials, the tab shows a message that there is no vulnerability data because the AI System has no bills of materials. Use Add Artifacts on the Overview tab to attach a Bill of Materials to see its vulnerabilities here.

When the AI System's Bills of Materials have no findings, the Models sub-tab shows "This AI System has no model vulnerabilities." and the Technologies sub-tab shows "This AI System has no open technology vulnerabilities."

If the AI System cannot be loaded, the tab shows "This AI System couldn't be loaded. Refresh the page to try again."

Managing Vulnerabilities

Security teams can resolve or ignore vulnerabilities based on remediation strategy. Resolving a vulnerability indicates it has been remediated, while ignoring acknowledges the risk but documents a decision not to address it at this time. Both actions require a free-text justification explaining your remediation steps or risk acceptance rationale, and the platform tracks which user performed each action for audit purposes. Vulnerability likelihood scores automatically recalculate when vulnerabilities are resolved or ignored.

When you remove a vulnerability, you attest that the detected vulnerability is not valid, and the platform attempts to ignore it in future scans.

Key Actions & Functionality

Resolve Model Vulnerabilities

Click Resolve Vulnerabilities in a model's Actions column to open the Remove Model Vulnerable Attack Categories pop-up. It lists every attack category found for the model, with its Severity, Remediation Guidance and a Reason for Removal field. Hover over the guidance to read the full text, or use the copy icon to copy it. Select Ignore or Resolve for each category, enter a reason, and click Yes. Remove these items. Clicking a severity tag or link in the Vulnerable Attack Categories column opens the same pop-up filtered to that severity.

Resolve or Ignore Technology Vulnerabilities

Click Ignore or Resolve in a technology's Actions column. To act on several at once, select their checkboxes and click Ignore Selected or Resolve Selected

View Model Details in the Arena

If a model has been tested in the AI Arena, its penetration test results appear on the Models sub-tab. Click the model's name to view the threat details in the Arena.

Restore Resolved Vulnerabilities

Restore previously resolved or ignored vulnerabilities if circumstances change. On the Technologies sub-tab, click Restore to open a list of resolved and ignored technology vulnerabilities, with the user who removed each one and the reason given. Select the items to restore and confirm. On the Models sub-tab, click the # Resolved link in a model's Vulnerable Attack Categories column.

Governance Information

The AI System Governance tab facilitates the management of governance and compliance for AI systems. It comprises four subsections:

  • System Information
  • Values
  • Data
  • Governance Contact

This tab ensures the AI system's purpose and intended use are explicitly defined, aligning its functionality with organizational goals and regulatory standards.

System Information

In the System Information section, you can modify the AI System's purpose, initial deployment date, last update date, and next planned update.

  • Purpose: Defined objective or intended application of the AI System.
  • Date of Initial Deployment: The date the AI System was first operational.
  • Last Updated: The most recent modification date of the AI System.
  • Next Regular Update: Scheduled date for the upcoming update of the AI System.

Values

The Values section is where you can update the AI system's value statement and provide a link to the organization's overarching values. The value statement helps align the system's goals with organizational principles.

  • Value Statement: A declaration of the organization's guiding principles and core values related to the AI System.
  • Link to Organizational Values: A URL or reference linking to the publicly available organizational values that guide the AI System's development and operation.

Data

In the Data section, users can document the data-related details of the system, such as any protected variables, the source of data (e.g., proprietary or open-source), and the levels of protection applied to the data.

  • Description: An overview of the data used by the AI system, including any sensitive or protected variables (e.g., gender, age).
  • Data Provenance: The original sources of the data (e.g., proprietary data, internal data, private data, open-source data, public data).
  • Data Protection: The levels of data protection applied (e.g., PII, PNII, financial and credit information, intellectual property, sensitive personal information, nonpublic personal information, material nonpublic information, classified information).

Governance Contact

The contact details of the individual responsible for governance oversight can be edited in the Governance Contact section, including the name, email, and title of the governance contact.

  • Contact Name: The name of the governance contact for the AI System.
  • Contact Email: The email address of the governance contact.
  • Position: The role or title of the governance contact within the organization.

AI System Settings

The AI System Settings tab equips users to manage and update foundational AI System details effectively. In this tab, users can update key system details, including the AI system name, description, and owner. These fields ensure proper system identification and attribution to the correct team or individual. Be sure to click the Save Changes button to save and edits made to these fields. If you navigate to another tab without saving, you will lose your changes.

Did this answer your question?