# Who Sells Data: data dictionary

Public, sourced directory of companies that sell AI training data, RL environments and evals, maintained by Secludy AI. Listings are free and unranked.

## directory.json / directory.csv

One record per company: `id`, `slug`, `profile_path`, `name`, `aliases`, `website`, `status` (active, acquired), `acquired_by`, `last_checked` (the date the company's facts were last checked).

Each fact (`facts.<field>` in JSON; `<field>`, `<field>__sources`, `<field>__checked` in CSV) has:

- `value`: the published value, or null.
- `status`: `listed`, `not_listed` (no public statement found; not a "No") or `not_confirmed` (sources don't settle it).
- `sources`: the public pages that support the value.
- `checked`: the date the value was checked.
- `quote` / `evidence`: the source's own words, kept for consequential claims (security, privacy, data provenance, named customers).
- `confirmed_by`: `vendor` or `person` when a fact was confirmed directly.

Fields: description, offering_types, domains, delivery_formats, public_samples, public_customers, hq, founded_year, headcount_band, latest_funding, total_funding, soc2, trust_center_url, other_certifications, deployment_options, data_sources, exclusivity, confidentiality_practices. Controlled values are in `taxonomy.json`. SOC 2 values starting "Vendor-stated" are the company's own statement, not an audit.

## Source policy

Ordinary company facts (what a company sells, where it's based, when it was founded, its size and funding) come from credible public sources: the company's own site and job posts, its Y Combinator or LinkedIn company page, reputable press, registries, and funding databases when nothing better exists. Claims about payments, eligibility, privacy and security are published only from the company's or program's own published words.

## claims.jsonl

History of published values: `claim_id`, `subject` (company id), `field`, `value`, `sources`, `observed_at`, `supersedes` (the claim it replaced). Use directory.json for current values.

## buying-programs.json

Programs that publicly pay organizations for their data, or pay bounties for introductions, as shown on the Who's buying page: company, program, kind, what they pay, what data they want, who qualifies, the program's own page and a verbatim quote, with the date observed. Amounts are each program's published terms; none is an offer.

Third-party material, including quoted passages and linked samples, keeps its own terms.