Introduction

Cranium Detect AI builds on current Code Sensor functionality by crawling your private VCS's code repositories to automate the discovery of repositories containing AI/ML-based technologies and models. Detect AI supports Python, JavaScript, TypeScript, and Java. This capability provides additional visibility into existing and developing AI-enabled projects to ensure organizations have an early inventory of all AI/ML usage and can detect potential Shadow AI usage. Detect AI provides additional oversight for an organization's AI stakeholders without relying solely on what a development team identifies, and without manually analyzing code to verify compliance.

Detect AI in the VCS Integrations List

The Integrations page provides a centralized interface for managing Detect AI scans. It displays a detailed list of VCS Integrations, including the Detect AI scan status (e.g., not started, queued, failed, success) and Detect AI scan results. You can manage VCS Integrations through the three-dot menu, which allows you to run a Detect AI scan and view Detect AI scan results.

Run Detect AI Scan

You can run a Detect AI scan on an integration from the action menu. Configurable parameters are also available when launching a scan from the New BOM Wizard:

• Project filter - select the projects to include in the scan
• Last commit date - skip any repositories with a last commit before the specified date
• Minimum # of commits - skip any repositories with less than the number of commits specified

View Scan Results

The Scan Results section provides a comprehensive overview of your Detect AI scan process, showing the current scan status, the date of the last scan, and a count of repositories scanned. This count is broken down into the number of repositories scanned or skipped based on the given parameters, out of the total repositories available under the integration. By hovering over this number, you can access a detailed breakdown of how many repositories were scanned, skipped, or failed. Additionally, the results of repositories with AI detected are displayed. The repository list includes project and repository names (linked to the actual repositories in your version control system), the last commit date, committer details (name and email), and the code owner (when the VCS supports repository owners). The scan status of each repository is displayed, with repositories marked as Skipped, AI detected, Failed, or Queued, and hovering over any of these statuses reveals more context about flagged parameters or errors. The page includes a blue "Run Detect AI Scan" button to trigger a new scan, a red "Abort Scan" button to cancel an ongoing scan (which can also be done through the action menu), and a search feature to find repositories by name. You can also export the scan results with the white "Export to CSV" button.