Getting Started with Veritensor

Veritensor is an open-source security scanner designed specifically for the AI Supply Chain. Unlike traditional SAST tools that focus on code, Veritensor focuses on AI Artifacts: Models, Datasets, and RAG Documents.

Why Veritensor?

Modern AI development involves downloading gigabytes of opaque binary files (Models) and processing untrusted documents (RAG). Traditional security tools are blind to these threats.

• Models: Is that .pkl file safe to load, or will it execute a reverse shell?
• RAG: Does that PDF resume contain white-text prompt injections?
• Data: Is your Parquet dataset poisoned with malicious URLs?

Veritensor answers these questions in seconds, locally, without sending your data to the cloud.

Installation

Veritensor is a Python CLI tool.

Basic Installation

For scanning models and notebooks:

pip install veritensor

Full Installation (Recommended) Includes support for RAG documents (PDF/Docx), Datasets (Parquet/CSV), and PII detection:

pip install "veritensor[rag,data,pii]"
python -m spacy download en_core_web_lg

Quick Start

1. Scan a Project

Navigate to your project folder and run:

veritensor scan .

Veritensor will recursively find all supported files (Notebooks, Models, Configs) and audit them.

2. Scan a Specific Model

Check a downloaded model before loading it:

veritensor scan ./llama-2-7b.bin --repo meta-llama/Llama-2-7b

This verifies the hash against the official Hugging Face registry and checks for malware.

3. Scan RAG Documents

Check a folder of PDFs before vectorization:

veritensor scan ./rag_documents --full-scan

What's Next?

• Explore the Threat Database to understand what we detect.
• Learn how to Secure your RAG Pipeline.
• Set up CI/CD Integration.