AIBOM · AI supply-chain transparency
Know what every AI system is built from
Inventory models, datasets, frameworks and external AI services in CycloneDX and SPDX, validate them against policy, and keep the record current as models change.
- Datasettxn-history-2019-24
- Modelcredit-scoring v3.1
- Frameworkxgboost 2.0.3
- Serviceloan-decision-api
- OwnerData Science · Risk
- GapDataset provenance missing
Why AIBOM
AI Bill of Materials
AI systems depend on models, datasets, frameworks and hosted services that change without a code release and rarely appear in a software SBOM. CERT-In's Version 2.0 guidelines define minimum AIBOM elements, and the EU AI Act requires high-risk providers to document third-party pre-trained components and training-data provenance. An AIBOM gives security, compliance and procurement teams one verifiable record of what each AI system contains.
What AIBOM tracks
Capabilities
What IntelliXBOM does with your AIBOM.
One governed inventory, correlated with the rest of your BOMs, the risks they carry and the controls they support.
Generate and ingest AIBOMs
Create AIBOMs or import them from suppliers and open-source generators in CycloneDX ML-BOM and SPDX 3.0 formats.
Validate against policy
Check each AIBOM against required-field policies such as the CERT-In AIBOM elements, and report missing fields per model and dataset.
Version history and diffs
Keep every AIBOM version and see exactly what changed after retraining, a new adapter or a provider switch.
Risk correlation
Correlate models, datasets and ML frameworks with vulnerabilities, known-exploited lists, licences and end-of-life data, and record VEX decisions.
Business context
Link each model and AI service to the applications and business services that depend on it.
Evidence and deployment
Map the inventory to framework controls with timestamped evidence, running on-premise, in private cloud or air-gapped.
How it works
From collection to evidence.
The same five-step loop runs continuously, so the AIBOM never becomes yesterday’s inventory.
- 01Discover
Collect models, datasets, frameworks and AI services from generated or supplier-provided CycloneDX and SPDX AIBOMs.
- 02Validate
Check each AIBOM against its schema and your required-field policy, and flag gaps such as missing hashes or data sources.
- 03Correlate
Match components against vulnerabilities, licences, EOL data and the business services that rely on them.
- 04Track change
Store each new AIBOM version with a diff so reviewers see what changed and when.
- 05Report
Map the inventory to framework controls and export timestamped evidence for audits and regulators.
Use cases
The questions AIBOM answers.
Each question resolves to a governed, versioned record, and to the frameworks that record helps provide evidence for.
Which of our services use a model fine-tuned from a base model that has just been reported as backdoored?
Lineage in the AIBOM links base models to fine-tunes and to the services that call them, so affected systems can be listed in one query.
Does this supplier's AIBOM meet CERT-In's minimum elements?
Policy validation checks every model component for the required fields and lists what is missing before the delivery is accepted.
What changed in our credit model since the last audit?
Version history shows each AIBOM revision with a diff of datasets, metrics, dependencies and artefact hashes.
Is a vulnerable ML framework version running under any production model?
Correlation matches framework versions in each AIBOM against vulnerability and known-exploited data and shows the affected models and services.
AIBOM resources
AIBOM guides, from fundamentals to procurement.
18 source-cited articles. Open the AIBOM resource hub →
Fundamentals
Tools & platforms
Operations
Comparisons
Compliance
Industries
AIBOM questions, answered
What is an AIBOM?
An AIBOM, or AI Bill of Materials, is a structured inventory of the components used to build, train and deploy an AI system: models, datasets, software frameworks, infrastructure and external AI services. It records versions, origins, licences and relationships so that risks can be traced to the systems they affect.
What does CERT-In require in an AIBOM?
CERT-In's Technical Guidelines Version 2.0, dated 9 July 2025, list minimum AIBOM elements in Table 10, beginning with model name, version, type, developer, licensing, software dependencies, ML models and algorithms, performance metrics, data source and data-set information. The guidelines are aimed at government, public-sector and essential-services organisations and software exporters.
Which formats support AIBOMs?
CycloneDX has supported machine-learning BOMs since version 1.5, using model components with a modelCard. SPDX 3.0, released in April 2024, added AI and Dataset profiles. Both are open standards and CERT-In names both formats.
Does the EU AI Act require an AIBOM?
The Act does not use the term, but its technical documentation for high-risk systems covers software versions, third-party pre-trained components and training-data provenance. After the AI Omnibus, high-risk rules apply from 2 December 2027 for Annex III systems and 2 August 2028 for Annex I products.
How is an AIBOM different from an SBOM?
An SBOM lists software packages and their dependencies. An AIBOM adds models, datasets, training lineage, performance metrics and intended use, and it changes whenever a model is retrained or a provider updates a hosted model.
Are there international minimum elements for an AIBOM?
Yes. In May 2026 the G7 Cybersecurity Working Group published Software Bill of Materials for AI: Minimum Elements, grouping recommendations into seven clusters from metadata to security properties. The guidance describes the clusters as not mandatory.
Sources
- Technical Guidelines on SBOM, QBOM & CBOM, AIBOM and HBOM, Version 2.0 (9 July 2025)CERT-In, Government of Indiawww.cert-in.org.in/PDF/TechnicalGuidelines-on-SBOM,QBOM&CBOM,AIBOM_and_HBOM_ver2.0.pdf
- AI Act Annex IV: Technical DocumentationEU AI Act Explorer (Future of Life Institute)artificialintelligenceact.eu/annex/4/
- AI Omnibus enters into force (27 July 2026)European Commissiondigital-strategy.ec.europa.eu/en/news/ai-omnibus-enters-force
- Introducing OWASP CycloneDX v1.5 (26 June 2023)OWASP CycloneDXcyclonedx.org/news/cyclonedx-v1.5-released/
- SPDX 3.0 release announcement (16 April 2024)Linux Foundationwww.linuxfoundation.org/press/spdx-3-revolutionizes-software-management-in-systems-with-enhanced-functionality-and-streamlined-use-cases
- Software Bill of Materials (SBOM) for Artificial Intelligence: Minimum Elements (May 2026)BSI with G7 Cybersecurity Working Groupwww.bsi.bund.de/SharedDocs/Downloads/EN/BSI/KI/SBOM-for-AI_minimum-elements.html
- Global Cyber Agencies Issue New SBOMs for AI GuidanceInfosecurity Magazinewww.infosecurity-magazine.com/news/new-sboms-for-ai-guidance-2026/
Sources checked in September 2026. Regulations and guidance change; always refer to the issuing body’s current publication. This content is for general information and is not legal advice.
The rest of the BOM Suite