Which buyer is best for documents only?

micro1 names documents most directly (SOPs, process documentation, internal knowledge bases). Scale AI lists documents and Confluence knowledge records inside a wider set. For a company that has closed, SimpleClosure takes Drive and Notion workspaces.

Can we talk to more than one buyer?

None of the pages we read forbids it before signing. Exclusivity, if any, is set in the agreement, so ask each buyer what its license grants before you sign one.

Do these buyers charge the company a fee?

Not on any of these four pages. Marketplaces are another matter: Datarade keeps a commission from the providers that list with it.

We print terms only after reading them on the company’s own site, and Mercor’s data pages would not open for us on November 7, 2026.

How often do these terms change?

Often enough that every figure here carries a date. micro1 altered its company referral terms at some point after June 2026, for example. Each change we spot goes in the changelog.

Who Licenses Company Documents for AI Training

Published terms by buyer

Published terms, as read on each buyer's own page
Buyer	What it lists	Published pay	Who qualifies	De-identification	Payment timing	Ownership
micro1 Company data partnerships	Names SOPs, process docs and internal knowledge bases outright, beside operations and legal records	Three tiers per company, starting at $100K and running past $3M	A full-time staff of 30 or more, mostly English material; US, UK and Canadian firms most sought	Handled by micro1, with representative samples open to your review	Paid within 30 days of acceptance (its FAQ instead says of invoicing)	You keep ownership
Scale AI Data Partnership	Files and documents that carry metadata, Jira and Confluence records, HR records stripped of identities	An illustrative $10K–$1M+ a partnership, rising with delivery cadence; Scale calls this no guarantee	Prefers companies of 40 or more people, yet looks at each case	Done on your own systems ahead of extraction; you sign off every package	Not published	Kept under your governance; ownership itself not addressed
Nyne Company data partnerships	Records more than documents: books of account, CRM, service and ticket history, Asana and Jira projects	Partnerships "commonly worth $100K or more"	Three years trading and a US payroll of 20+ full-time W-2 staff	Released only once de-identified; whose job that is goes unsaid	Not published	Set by the agreement you sign
SimpleClosure AssetHub	Shutdown-stage workspaces (Slack, Drive, mail, Notion, Figma) alongside code and IP	Not published	Founders closing their company through SimpleClosure; no written criteria	Each dataset gets a personal-data pass ahead of any license; who runs it is unstated	Not published	Not published

Disclosure: several of these buyers reward introductions that end in a signed license. Rows are ordered by how much each page says about documents, and a fee moves nothing here.

Independence note: no company on this page owns, backs or has approved this site.

What does each buyer's page actually say?

micro1

Three bands. The entry band, $100K to $500K, is for what micro1 calls qualified partnerships. The middle, $500K to $3M, needs a large operational dataset or several teams taking part over time. Above $3M is reserved for operational data it considers highly unique. The bar to enter: an established firm with at least 30 people on full-time payroll, records chiefly in English, ideally based in America, Britain or Canada. micro1 runs de-identification itself and may rewrite some sets synthetically; you can inspect representative samples, and the data remains yours. Payment timing is stated twice and not quite the same way: 30 days from acceptance in one place, 30 days from invoice in the FAQ.

Scale AI

Scale's figure, $10K up to $1M+, comes with two caveats on the page: it is illustrative, and it grows with how often you deliver. Forty people is where it would like to start, though smaller firms get looked at individually. Its wish list covers documents and files together with their metadata, Jira and Confluence content, and HR records once identities are gone. Credentials, raw system access and anyone's identity are ruled out. The anonymizing is done on your side of the wall, before extraction, and no package moves without your approval.

Nyne

Nyne's own phrase for a typical partnership is "$100K or more," with the figure turning on what data you have and how much of it. To qualify, a company needs three years of trading behind it and twenty or more full-time W-2 employees in the US. What it lists is system data: books of account, CRM, service and ticket history, Asana and Jira projects. Documents count here only when they live inside those tools. Its page says data is de-identified ahead of release.

SimpleClosure AssetHub

The route for a company that is closing. Founders who wind down through SimpleClosure can offer their Slack, Drive, mail, Notion and Figma workspaces, along with code, IP and domains. No pay figures appear on the page. It does say each dataset gets processed for personal data ahead of licensing.

Who did we check and leave out?

Mercor. Our November 7 attempt to open its data pages failed, so we show none of its terms.
Labelbox Prism. Prism asks for reasoning traces, support logs and robot telemetry rather than internal documents. Worth knowing if you hold those: its licenses default to exclusive, and it claims under 21 days from a first call to money wired.

Marketplaces work the other way round: the seller writes the listing and sets the price. Datarade's provider plans run from a free tier that keeps 30% commission to a $12,000-a-year plan that keeps 15%. AWS keeps 3% of public Data Exchange offers as its listing fee. Neither one buys anything itself, which is why they sit outside the table.

What changed in buyers' terms?

Terms changelog, newest first.

Nov 7, 2026: Baseline: we opened each buyer page on this site and copied its terms. Later entries record what moves.
Nov 7, 2026: micro1: its data page currently shows per-company tiers topping out at $3M+.
Nov 7, 2026: micro1: the company referral page sets $50,000 per deal worth up to $700,000; past that, it pays 7.5%. The June 16, 2026 announcement gave different terms; no change date is given.
Nov 7, 2026: Mercor: pages unreadable on this date. It joins the table once we can read them firsthand.

Sources, with read dates

micro1.ai/data (opened 11/7/2026)
micro1.ai/company-referral (opened 11/7/2026)
micro1 forum post on company referrals, June 16, 2026 (opened 11/7/2026)
scale.com/data-partnership (opened 11/7/2026)
nyne.ai/sell-company-data (opened 11/7/2026)
simpleclosure.com/product/assethub (opened 11/7/2026)
prism.labelbox.com (opened 11/7/2026)
providers.datarade.ai/apply (opened 11/7/2026)
AWS Marketplace documentation, listing fees (opened 11/7/2026)

Questions on file

Which buyer is best for documents only?: micro1 names documents most directly (SOPs, process documentation, internal knowledge bases). Scale AI lists documents and Confluence knowledge records inside a wider set. For a company that has closed, SimpleClosure takes Drive and Notion workspaces.
Can we talk to more than one buyer?: None of the pages we read forbids it before signing. Exclusivity, if any, is set in the agreement, so ask each buyer what its license grants before you sign one.
Do these buyers charge the company a fee?: Not on any of these four pages. Marketplaces are another matter: Datarade keeps a commission from the providers that list with it.
Where is Mercor?: We print terms only after reading them on the company’s own site, and Mercor’s data pages would not open for us on November 7, 2026.
How often do these terms change?: Often enough that every figure here carries a date. micro1 altered its company referral terms at some point after June 2026, for example. Each change we spot goes in the changelog.

Who licenses company documents for AI training?