Foundations & Open-Source StewardsPowered by Anchor

The Custodian Model

How an open-source steward turned a chaotic fork forest into a certified model lineage.

This story sits inside our mission to make digital provenance effortless, universal, and trustworthy for the people and systems who depend on it. See Mission & Vision

Publish your certified model lineage See how Sovereign maps model trees

Problem … When everyone ships "your" model

OpenFoundation releases a base model: ethically framed, documented, with clear license and use guidelines. Within months: fine-tuned versions proliferate on GitHub, Hugging Face, internal forks. Some keep the foundation's name. Some violate the spirit or letter of the license.

To the public: it is impossible to tell which versions are actually aligned with the foundation's standards.

Summary:

No canonical lineage tree.
No "blessed" list vs "unreviewed" list.
Reputation risk when misaligned derivatives claim their name.

Solution … A public, anchored model family tree

Step 1: Register the base model

They register the base model as a root artifact with a model passport that includes: license, training data policies, governance notes, version and commit hashes, sovereign zone. The passport is hashed, anchored externally (OTS / IPFS), given a stable public ID.

Step 2: Define a "certified lineage" program

OpenFoundation publishes how downstream projects can request certification, what must be shared: dataset passports, training modifications, evaluation results, safety constraints. Each accepted downstream model is registered in Sovereign as a child of the base model, gets its own model passport, is linked with a relation type such as "fine_tune", "distillation", "adapter", etc.

Step 3: Public lineage explorer

Sovereign provides a public view: base model at the root, certified descendants clearly labeled, non-certified / unknown forks outside their tree.

Flow … How a downstream model becomes certified

Step 1: Request

A downstream team submits a certification request. They share references to their dataset passports, training run IDs, evaluation notes.

Step 2: Review

OpenFoundation reviews: licensing compliance, training data alignment with policies, safety / evaluation results.

Step 3: Certification

If accepted: they register the downstream model as a child in the model lineage. Sovereign updates the tree marking this node as "Certified by OpenFoundation." Anchors the updated lineage.

Step 4: Public display

The downstream team can now link to a public SOVEREIGN\PROVENANCE page showing: "Certified descendant of [Base Model] by OpenFoundation."

Outcomes … From name chaos to a verifiable tree

The foundation sets a de facto standard for what "officially derived from us" means, protects its name by separating certified from non-certified forks.
Downstream teams gain credibility by being visibly on the certified tree.
Policymakers, journalists, and users have a canonical reference for which models are governed by the foundation's policies, what changes were made along the lineage.

"Instead of arguing on social media about 'our model' vs 'their model,' we point to the tree."

Product tie-in … Why this is SOVEREIGN\\PROVENANCE

Sovereign gives models passports, connects them in a graph, anchors that graph. It is not just metadata in a repo, not just a README badge. It is a cryptographic lineage that can be cited in papers, linked in model cards, referenced in policy docs.

If you steward a base model, you should steward its lineage too.

Publish a certified lineage for your models.

SOVEREIGN\PROVENANCE turns your stewardship into a public, verifiable tree.

Publish your certified model lineage Talk to us about model stewardship