# Closed model AIBOM: checklist, procurement questions and clause starting points

Companion to the Safeguard post "Why Anthropic and OpenAI Do Not Publish AIBOMs for Their Models" (23 October 2026). This is a working template, not legal advice. Have your counsel adapt any clause before use.

## Part 1: record each closed model you use

For each model, per environment:

- [ ] Every call site found (code, configuration, environment variables, gateways) and owned by a named team
- [ ] Exact model ID recorded (for example the API ID, plus the cloud platform ID if you call it through a cloud platform)
- [ ] Identity fields transcribed from the provider's own pages, each with URL and the date you read it: supplier, release date, knowledge or training data cutoff, modalities, context window, retirement date where given
- [ ] Links to the system card and model documentation added as external references
- [ ] Training data entry quotes the provider's own category description, with source URL
- [ ] Training data composition marked unknown in a machine readable way (CycloneDX: `compositions` entry with `"aggregate": "unknown"`; SPDX 3: "trainedOn" relationship with `completeness` set to `noAssertion`)
- [ ] EU AI Act public training content summary located and linked, or recorded as not located (with the date you checked)
- [ ] Your data flows recorded: data classes sent, provider account, region, data usage setting for training, retention controls
- [ ] Your own evaluation evidence attached (accuracy on your test set, refusal behaviour, prompt injection tests)
- [ ] Risk acceptance written: owner, basis, use case scope, expiry date
- [ ] Review trigger set: model ID change, new system card, new EU summary, or every six months, whichever comes first

## Part 2: questions to ask any closed model provider

1. Which model IDs and versions does this agreement cover, and how will you notify us of new versions and retirements?
2. Will you provide the EU AI Act public training content summary for each model, and the Annex XII documentation if we integrate the model into an AI system?
3. Can you describe the training data categories for this model in more detail under NDA, including whether licensed datasets were used and in which domains?
4. How do you handle reservations of rights and opt outs for web data, and how can a rightsholder check?
5. What personal data filtering do you apply, and how do you handle data subject requests that relate to training data?
6. Is any of our data used for training under this agreement? What are the default and optional retention periods?
7. Can you supply a machine readable model card or AIBOM, in CycloneDX or SPDX, for this model? If not now, is it planned?
8. What evaluation results can you share for our use case or domain beyond the public system card?

Record every "no" or "not available" answer as a documented unknown in your AIBOM.

## Part 3: clause starting points

- **Documentation.** For each covered model version, the provider will supply the public system card, the EU training content summary where it exists, and any Annex XII documentation, within [N] days of release.
- **Change notification.** The provider will notify the customer at least [N] days before retiring or materially changing a covered model ID.
- **Data use.** Customer content will not be used to train or improve models without written opt in. Retention periods: [state].
- **Regulatory assistance.** The provider will reasonably assist with regulatory inquiries about the model's training data, within the limits of trade secret protection.
- **Machine readable documentation.** Where the provider produces machine readable model documentation, it will make it available to the customer.

## Primary sources

- EU AI Act Article 53: https://artificialintelligenceact.eu/article/53/
- Commission template for the public summary of training content (24 July 2025): https://digital-strategy.ec.europa.eu/en/library/explanatory-notice-and-template-public-summary-training-content-general-purpose-ai-models
- General Purpose AI Code of Practice: https://digital-strategy.ec.europa.eu/en/policies/contents-code-gpai
- CycloneDX ML-BOM: https://cyclonedx.org/capabilities/mlbom/
- SPDX 3.0.1 AI profile: https://spdx.github.io/spdx-spec/v3.0.1/model/AI/AI/
- NIST AI 600-1: https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence