An AI SBOM is a machine-readable inventory of a model, its training datasets, and their provenance and licenses.
- Where a software SBOM carries package dependencies, an AI SBOM also covers what a software SBOM leaves out: model weights and training data.
- It is built from what the model card says, so a thin model card leaves an AI SBOM with that many empty entries.
Why produce one
It serves two purposes.
- For the publisher it is a self-check on how complete the documentation is. Seeing what is missing as a list means you can close the gaps before release.
- For the recipient it is evidence for a review. Just as SK Telecom asks suppliers for an SBOM, other organizations have begun asking the same of AI models.
Key regulations
Nothing on this page certifies or determines compliance with any regulation.
- It makes documentation gaps visible so a person can prepare.
- Interpreting them against a specific system’s legal obligations is a person’s job; when in doubt, consult Legal and the OSRB.
| Regulation | Effective | Key provisions |
|---|---|---|
| EU AI Act | High-risk and transparency duties from 2 August 2026 | Article 11, Annex IV |
| Korea’s AI Framework Act | In effect since 22 January 2026 | Article 31 (transparency), 32 (safety), 33–34 (high-impact AI), 35 (impact assessment) |
- The EU AI Act is specific about technical documentation through Article 11 and Annex IV: it must cover a system’s purpose and architecture, the provenance and processing of its training data, and its performance and limits.
- Korea’s AI Framework Act keeps its detailed documentation requirements in the enforcement decree, so it is less specific than the EU’s. Article 32 (safety) applies only to systems trained above a compute threshold set by that decree, so a linked element points at the subject of the duty rather than establishing that the duty applies.
The G7 minimum elements
In May 2026, CISA and G7 partners jointly published “Software Bill of Materials for AI — Minimum Elements”, led by Germany’s BSI and Italy’s ACN. It defines 50 minimum elements in seven clusters that a model’s inventory should carry. It is a non-binding recommendation, not a regulation.
| Cluster | Elements | Needing human judgement | Content |
|---|---|---|---|
| Metadata | 10 | 0 | Who produced the inventory, when, with which tool |
| System-level properties | 9 | 4 | System name, application area, data flow |
| Models | 13 | 0 | Model identifier, license, integrity, training properties |
| Dataset properties | 10 | 5 | Dataset provenance, statistics, sensitivity, license |
| Infrastructure | 2 | 0 | Software dependencies and hardware |
| Security properties | 4 | 3 | Security controls, policy, vulnerability handling |
| Key performance indicators | 2 | 1 | Security metrics and operational performance |
| Total | 50 | 13 |
- Thirteen of the 50 have no automated source. Things like the intended application area or the sensitivity of the training data cannot be proven by any model card field, so a person has to supply them.
- The documentation Annex IV of the EU AI Act asks for corresponds to the G7 system-level, model, and dataset clusters.
- Advisory as they are, the elements overlap substantially with the documentation those regulations require. Filling the G7 side also builds most of what a regulatory submission needs.
Producing an AI SBOM
Give BomLens, SK Telecom’s open-source SBOM generator, a model id and it reads the model card, builds a CycloneDX AI SBOM, and reports G7 element coverage alongside the regulatory mapping. (It fetches model-card metadata only and does not download the weights.)
- Setup, usage and how to read the reports are covered in the BomLens AI model guide.
- The command below assumes BomLens is already installed.
./scripts/scan-sbom.sh --project my-llm --version 1.0.0 \
--model "my-org/my-llm" --generate-only
A private repository needs a token with read access.
- How to pass it as
HF_TOKEN, and what a gated repository additionally requires, are in Private and gated models.
You get:
- an AI SBOM covering the model and its datasets
- per-element coverage, and how to supply what is missing
- the mapping to the EU AI Act and Korea’s AI Framework Act
- components whose license needs human review
Checking the result
An overall pass does not mean every G7 element is filled. G7 elements are all advisory and do not move the verdict. Read the covered count and the gap count separately.
- Gaps come in two kinds: those you close by writing in the model card, and those a person has to judge. Start with the first kind.
- To see what the reports look like before running anything, open the conformance report and the AI compliance profile from a sample model scan.
Related pages
- Model card — the input to an AI SBOM
- Pre-release checklist — what to confirm before release
- What is an SBOM — SBOMs in general
- BomLens AI model guide — setup, usage, reading the reports
- BomLens — the tool, and its supplier-facing usage
References
- Software Bill of Materials for AI — Minimum Elements (G7, May 2026)
- Regulation (EU) 2024/1689 (AI Act)
- AI Framework Act (Korea)
Contact
For questions about producing or reading an AI SBOM, contact the OSRB (opensource@sktelecom.com).