AI Inventory: Managing Risk from AI-Generated and Imported Models

Hopper’s AI Inventory helps you track and manage security and compliance risk introduced by AI models in your codebase. Whether models are generated, imported, or modified, Hopper surfaces risk based on behavior, metadata, and usage patterns.

AI Inventory View Fields

Each model entry includes:

• Project (where the model lives)
• Model ID
• Risk Classification (Malicious, Remote, Safe)
• Pickle (presence of insecure deserialization)
• Model Source (e.g. Hugging Face, internal)
• Download Count
• Sensitive Data (if detected)
• License Type
• Usage (with direct link to the model in your SCM)

AI-Specific Filters

Use the following filters to surface AI model risk:

• Show Crown Jewels Projects only
• Show Projects with Malicious Models only
• Show Remote Projects only
• Model Source
• Integration
• Download Count: Greater than / less than
• Crown Jewels and Custom Tags

These filters help you identify not just vulnerable OSS but also risky models that may expose your systems to supply chain or licensing risks.

Exporting AI Model Risk Data

Filtered AI Inventory results can also be exported as CSV or JSON format. This data supports compliance reviews, engineering audits, and automated tooling.