Contribute

The catalog is a set of plain YAML files in a Git repository. Every change — a new entry, a correction, a policy revision — is a pull request that anyone can read, and every pull request is checked by the same validation the site uses to build.

Report a correction

If something is wrong or out of date, the most useful report includes the entry, the specific statement, and a source that supports the correction (an official page, license file, model card, repository, or filing). Each entry page has Report a correction and Edit this entry actions that open the right issue form or file once the repository is configured. Corrections are handled the same way for everyone, including listed organizations.

Suggest a new entry

Use the entry-request issue form, or open a pull request with a new record. Before suggesting an entry, check the eligibility policy: the entry needs a documented U.S. basis, and for projects that basis is the governing or maintaining entity, not the contributors. Requests that cannot yet be verified are kept in the internal review queue (CONTENT_REVIEW.md) rather than published.

Editing workflow

1. Fork the repository and create a branch.
2. Add or edit a file in content/organizations/ or content/artifacts/. The file name must match the record's slug. The full field reference is in docs/CONTENT_FORMAT.md.
3. Cite a source for every statement. Only cite pages you have opened and read, and record the date you read them in accessed_at.
4. Set last_reviewed to today only if you checked the entry's evidence. Set updated_at to today for any substantive edit. Never change released_at without a source.
5. Run npm run validate. It checks schemas, unique slugs and source IDs, relationships between records, source references, eligibility and publication consistency, dates, family and release structure, and placeholder text.
6. Open a pull request using the template. A maintainer reviews the evidence before merging.

Evidence requirements

• Licenses come from the license file or the model or dataset card, not from a summary on another site.
• Headquarters come from the organization's own pages or, for public companies, regulatory filings.
• News reporting is acceptable only for the specific claim it supports, and is labeled as a news report.
• Do not add funding, valuations, staff counts, user counts, benchmark scores, hardware requirements, or power figures unless the catalog's rules for that field are met. Most of these are not published at all.
• If a fact is unknown, leave it unknown.

Sample organization record

This abbreviated example uses a fictional organization to show the shape of a record. It is not catalog content.

slug: sample-lab
name: Sample Lab
legal_name: null
website: https://sample-lab.invalid
logo_text: SL
summary:
  text: Sample Lab develops language models and publishes some of them with open weights.
  source_ids: [about]
organization_roles: [model-developer, research-lab]
ownership_category: privately-held
headquarters:
  label: Denver, Colorado
  country: US
  source_ids: [about]
products:
  - id: sample-api
    name: Sample API
    kind: hosted-model-api
    access: Paid API access to hosted models, documented in the developer docs.
    url: https://sample-lab.invalid/api
    source_ids: [api-docs]
    last_reviewed: 2026-09-29
eligibility:
  status: eligible
  basis: us-headquarters
  explanation: The company's about page gives its headquarters as Denver, Colorado.
  source_ids: [about]
  assessed_at: 2026-09-29
publication_status: draft
updated_at: 2026-09-29
last_reviewed: 2026-09-29
sources:
  - id: about
    title: About Sample Lab
    url: https://sample-lab.invalid/about
    publisher: Sample Lab
    kind: official-page
    published_at: null
    accessed_at: 2026-09-29
  # …plus the api-docs source

Sample model-release record

slug: sample-model-1-7b
name: Sample Model 1 7B
kind: model
record_level: release
family_slug: sample-model
version: "1 (7B)"
maintainers:
  - name: Sample Lab
    organization_slug: sample-lab
    source_ids: [model-card]
organization_slugs: [sample-lab]
summary:
  text: A 7-billion-parameter language model released with downloadable weights.
  source_ids: [model-card]
availability:
  status: public
  access_conditions: Downloadable from the model hub without an access request.
  source_ids: [model-card]
licenses:
  - name: Apache License 2.0
    spdx: Apache-2.0
    url: https://sample-lab.invalid/LICENSE
    applies_to: weights-and-code
    source_ids: [license]
checklist:
  weights: { status: public, note: null, source_ids: [model-card] }
  training_code: { status: unknown, note: null, source_ids: [] }
released_at: 2026-05
# …eligibility, dates, and sources as above

The openness tier is never written by hand: the site computes it from the checklist and licenses using the published rubric.

Changing the rules

Changes to the eligibility policy, the openness rubric, or the USASI Compact are proposed the same way, by pull request, and must update ELIGIBILITY.md, OPENNESS.md, or the Compact together with the code that applies them. Rubric changes get a new version number and a changelog entry.

Code contributions

The site is Next.js with TypeScript, Tailwind CSS, and Zod, exported as static files. Run npm run typecheck, npm run lint, npm test, and npm run test:e2e before opening a pull request. The README describes local development and deployment.