tibet-ai-sbom 0.1.0

BSI/G7 SBOM-for-AI implementation — software, models, datasets, infrastructure, security, KPIs. With TIBET provenance + CBOM evidence linkage.

tibet-ai-sbom

BSI / G7 SBOM-for-AI implementation, on top of TIBET provenance.

SBOM answers: what is present. CBOM answers: how it got here and what happened to it. This package builds the AI-SBOM document. CBOM/TIBET packages provide the causal evidence beneath it.

What this package is

tibet-ai-sbom implements the Software Bill of Materials for AI — Minimum Elements specification published by the German Federal Office for Information Security (BSI) in cooperation with G7 partners.

It is the first PyPI package to address the BSI AI-SBOM expectations as a first-class concern (as of 2026-05-15). The package takes BSI's seven clusters — Metadata, System Level Properties, Models, Dataset Properties, Infrastructure, Security Properties, Key Performance Indicators — and exposes them as a stable set of cluster codes, CVE-style indexable.

This 0.1.0 is the honest foundation:

• the BSI cluster codes are exposed and indexable
• the conformance roadmap is published openly (see ROADMAP.md)
• the workspace-scan entry point is in place as a placeholder
• coverage status per cluster element is honest, not aspirational

Full coverage of Models, Datasets, and KPIs follows in subsequent releases.

Why this exists

A normal SBOM tool answers: which dependencies are present. An AI system is more than that. An AI system spans:

• many sibling packages in one workspace
• one or more model artifacts and their training provenance
• supporting datasets and their sensitivity classes
• runtime infrastructure including accelerators
• security properties and AI-specific controls
• operational KPIs including drift

Auditors and procurement officers reading the BSI paper need a single place to map those expectations onto a real package. That is what tibet-ai-sbom provides.

Cluster codes (CVE-style indexable)

Every BSI minimum element is addressable by a short, grep-able code.

Code prefix	Cluster
AISBOM-MD-	Metadata
AISBOM-SLP-	System Level Properties
AISBOM-MOD-	Models
AISBOM-DSE-	Dataset Properties
AISBOM-INF-	Infrastructure
AISBOM-SEC-	Security Properties
AISBOM-KPI-	Key Performance Indicators

Example: AISBOM-MD-001 refers to the SBOM author element of the Metadata cluster.

This convention is deliberately CVE-style (CVE-YYYY-NNNN) so engineers and auditors can refer to a single specific requirement by code rather than by paragraph.

Install

pip install tibet-ai-sbom

The generic alias ai-sbom is provided for discovery and is kept in lock-step with tibet-ai-sbom:

pip install ai-sbom        # = same package, pinned to tibet-ai-sbom

Quick start

# List all cluster codes
tibet-ai-sbom clusters

# Filter by cluster
tibet-ai-sbom clusters --cluster MOD

# Describe a single code
tibet-ai-sbom code AISBOM-MD-003

# Workspace scan placeholder (full impl on the roadmap)
tibet-ai-sbom scan /path/to/workspace

Coverage today

Cluster	Status
Metadata	partial
System Level Properties (SLP)	partial / weak
Models	missing (planned)
Dataset Properties (DSE)	missing (planned)
Infrastructure	partial
Security Properties	partial
Key Performance Indicators	missing (planned)

Honest version of this table: see CONFORMANCE.md. Plan to close the gaps: see ROADMAP.md.

Where this fits in the broader stack

┌─────────────────────────────────────────────────────────────┐
│ tibet-ai-sbom        AI-SBOM overlay schema (this package)  │
├─────────────────────────────────────────────────────────────┤
│ tibet-sbom           Software SBOM + provenance (substrate) │
├─────────────────────────────────────────────────────────────┤
│ tibet-cbom           Continuity Bill of Materials (causal)  │
│ tibet-keychain       Custody and chain walk                 │
│ tibet-trail          Audit trail / search / verify          │
│ tibet-twin           Drift and operational state            │
│ tibet-continuityd    Sealed handoff and continuation        │
├─────────────────────────────────────────────────────────────┤
│ TIBET core           Identity-bound, causally ordered       │
│                      provenance substrate                   │
└─────────────────────────────────────────────────────────────┘

The AI-SBOM document is produced by this package. The evidence beneath it is provided by the wider TIBET / CBOM family. The two layers are linked explicitly through evidence references — not by embedding causal history into the SBOM file itself.

Reference

This package follows the cluster structure of:

Software Bill of Materials for AI — Minimum Elements, Bundesamt für Sicherheit in der Informationstechnik (BSI), in cooperation with G7 partners, 2026.

See the official source at BSI for the authoritative paper.

Status

• Version: 0.1.0 (alpha)
• License: MIT
• Stability: API may evolve as the BSI specification evolves and as cluster coverage grows.

Authors

• Jasper van de Meent · Humotica
• Root AI (Claude) · Humotica

One love, one fAmIly!