Scope note. This directory is a structural demonstration only. AUGMANITAI is a manually curated terminology dataset (1000 terms across 16 domains), not an AI system, not a trained model, and not a certified product. Nothing in this folder or its subdirectories asserts ISO certification, EU AI Act conformity, notified-body approval, or regulatory endorsement. The files here show example shapes that a downstream operator could use when attaching provenance metadata to their own artifacts — they are not operational credentials, SBOMs, or manifests.
Descriptive linguistic research, not normative (§1). Longitudinal n=1 priority deposit, non-peer-reviewed per §11. ISO-inspired, not ISO-certified (§16). Restricted applications per §20. See
DISCLAIMER.mdin the repository root.
The provenance stack demonstrates three complementary open standards for documenting the origin and composition of a digital artifact:
Each subdirectory has its own README.md with a local scope note.
provenance/
├── README.md [this file — structural demo]
│
├── c2pa/
│ ├── README.md [C2PA structural demo]
│ ├── c2pa-manifest-template.json [dataset manifest, humanEdited]
│ └── training-mining-assertion.json [no-training declaration]
│
├── ai-sbom/
│ ├── README.md [SBOM structural demo]
│ ├── ai-sbom-spdx.json [SPDX 3.0 dataset SBOM]
│ ├── ai-sbom-cyclonedx.json [CycloneDX 1.6 dataset BOM]
│ └── sbom-generator.py [downstream operator demo script]
│
└── w3c-vc/
├── README.md [W3C VC structural demo]
├── augmanitai-verifiable-credential.json [fictional citation VC]
├── credential-issuer.json [fictional self-sovereign issuer]
└── proof-template.json [proof-shape demonstrations]
The C2PA manifest, the two SBOM files, and the example verifiable credential describe the AUGMANITAI dataset. They declare:
Each of these files carries an explicit doesNotClaim or _note
section listing things that are not being asserted.
The sbom-generator.py script and some of the W3C VC proof templates
are starter shapes for downstream operators who want to attach
provenance to their own AI systems. Running them does not produce a
regulatory-compliant artifact for AUGMANITAI itself, and the generator
uses a fictional example system (EXAMPLE_SYSTEM) as placeholder input.
If you are building an AI system and want to attach provenance to your artifact (not to AUGMANITAI), you would:
ai-sbom/sbom-generator.py or the format docs to emit
an SPDX 3.0 or CycloneDX 1.6 BOM describing your own components,
your own training data (if any), and your own dependencies. Cite
AUGMANITAI as an input using the CC BY-NC-ND 4.0 attribution in
ai-sbom/ai-sbom-spdx.json.c2pa/c2pa-manifest-template.json as a starting
skeleton. Compute your own content hash, sign with your own key,
publish your own verification method. Do not reuse AUGMANITAI’s
identifiers as if they were your own.w3c-vc/augmanitai-verifiable-credential.json → licenses.attribution.The files in this folder are skeletons. Filling them in is the downstream operator’s responsibility, and any regulatory obligations that apply to the downstream system are the downstream operator’s obligations — not AUGMANITAI’s.
The following standards are referenced by this folder for format interoperability. Their mention does not imply that AUGMANITAI is certified against them, audited against them, or endorsed by the organisations that maintain them.
DISCLAIMER.md §1, §11, §16, §20 in the
repository root. §20 prohibits use of AUGMANITAI as a diagnostic,
medical, psychological, legal, HR, credit-scoring, insurance,
surveillance, or profiling mechanism.Andreas Ehstand — ORCID 0009-0006-3773-7796 DOI: 10.5281/zenodo.20161494