GitHub Repository

model-due-diligence is not a model safety verifier. It is a static evidence-gathering control for AI model supply-chain review. It supports provenance, artefact integrity, unsafe serialisation detection, secret exposure checks, suspicious code review, dependency risk detection, and audit reporting before first model execution.

1 starsPython

Model Due Diligence

Name: Model Due Diligence
Availability: InStock
Author: djhope99

by djhope99·Jun 13, 2026·3 points·0 comments

Visit Project View on HN

AI Analysis

●●SolidBig BrainNiche Gem

Aggregates ModelScan, Semgrep, and Bandit for AI model supply-chain review.

Strengths

•Checks pickle, safetensors, and GGUF formats for unsafe serialization.
•Detects exposed secrets and suspicious AST patterns like eval and exec.
•Produces deterministic JSON reports suitable for CI/CD automation pipelines.

Weaknesses

•Zero stars and forks suggests very early adoption and untested in production.
•ModelScan already handles unsafe serialization detection independently.

Post Description

model-due-diligence is a Python command-line tool for performing static evidence-gathering control for AI model supply-chain review. It supports provenance, artefact integrity, unsafe serialisation detection, secret exposure checks, suspicious code review, dependency risk detection, and audit reporting before first model execution.