Skip to main content

The AI Integrity eXchange & Verification Protocol.

Project description

AIXV: AI Integrity eXchange & Verification

An open protocol for AI supply chain provenance.

AIXV brings transparency and trust to AI systems by enabling cryptographic verification of:

  • model lineage (who built and modified this artifact),
  • training data provenance, and
  • safety and alignment attestations.

Features

  • Provenance Verification: Verify that a model, dataset, or pipeline artifact matches a signed AIXV manifest, including lineage and declared usage.
  • Integrity Checks: Hash- and signature-based verification to detect tampering between build and deployment.
  • Transparency Ledger (Coming Soon): Optional integration with the AIXV Transparency Ledger for immutable, queryable records of AI artifacts and attestations.
  • CLI Reference Implementation: A simple aixv CLI for developers and researchers to create and verify manifests, designed to evolve into the canonical verifier.

Threat Model

AIXV is designed to let deployers verify that the model artifact they are about to run is exactly the one their vendor or internal pipeline attested to, even if storage or transport is compromised. It protects against:

  • Malicious Model Substitution: Ensuring the model hasn't been swapped for a backdoored version.
  • Tampering: Detecting unauthorized modifications to weights or configuration.

Ecosystem

AIXV is designed to sit alongside software supply-chain standards like SLSA and Sigstore, but is tailored to AI artifacts (models, datasets, prompts, safety attestations). It focuses on the specific metadata and verification needs of the AI lifecycle.

Installation

You can install aixv directly from PyPI:

pip install aixv

Usage

The aixv CLI provides commands to interact with the protocol.

Verify a Model

To verify a model artifact against its AIXV manifest:

aixv verify model.pt --manifest model.aixv.json --proof proof.json

Note: The reference implementation currently focuses on local verification and proof object simulation for experimentation. Do not use it yet for production security guarantees.

Check Version

To check the installed version of aixv:

aixv version

Development

To set up the development environment:

  1. Clone the repository:

    git clone https://github.com/aixv-org/aixv-core.git
    cd aixv-core
    
  2. Install dependencies:

    pip install -e .
    

Status

This reference implementation is Pre-Alpha and intended for experimentation only.

  • The protocol is still evolving.
  • Connectivity to the AIXV Transparency Ledger is currently restricted to partner nodes.

For documentation and protocol specifications, visit aixv.org.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

aixv-0.0.1.tar.gz (7.8 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

aixv-0.0.1-py3-none-any.whl (7.3 kB view details)

Uploaded Python 3

File details

Details for the file aixv-0.0.1.tar.gz.

File metadata

  • Download URL: aixv-0.0.1.tar.gz
  • Upload date:
  • Size: 7.8 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.9.6

File hashes

Hashes for aixv-0.0.1.tar.gz
Algorithm Hash digest
SHA256 afc94b65a66412c5ba316a7f13130e391cd1d670aadbd151b891eb253371b9a4
MD5 4b1f71ced9c08d64173958189655b63c
BLAKE2b-256 f90d265acc5049f7d16b6f69d13ab2b93989bef862724b8e348e29cbe437f07d

See more details on using hashes here.

File details

Details for the file aixv-0.0.1-py3-none-any.whl.

File metadata

  • Download URL: aixv-0.0.1-py3-none-any.whl
  • Upload date:
  • Size: 7.3 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.9.6

File hashes

Hashes for aixv-0.0.1-py3-none-any.whl
Algorithm Hash digest
SHA256 1da49895c1f97e44d61977971c0dadae01fc0c10a1cfaebb6c0c3917882da3fb
MD5 4ef2610ed9f3096dc6c88925f5c411a8
BLAKE2b-256 fb08e6dcf714bbc7506123bb18283ee0737881dfa627e6b3abdbb86cc5849eee

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page