Data sale readiness is whether a company meets the entry thresholds AI data programs publish and can license usable records it has the right to license, with the people in place to sign and perform. General headcount floors run from 20 to 50 employees, with lower sector floors at Mode, each the program's own statement, checked November 9, 2026.
Readiness is the first question in AI training data licensing. Meeting a headcount floor is an entry rule and not an offer, export windows can close within 60 to 180 days, and not proceeding is a valid result of the check.
DataPacts Editorial Team checked every threshold here against the program's own page on November 9, 2026; the page on how we source, date and check every figure sets out the rules.
What Data Sale Readiness Means Before an AI Data License
Readiness has four parts, and a company can pass the first and still fail the other three.
- Eligibility: does the company meet the program's entry rule?
- Data: does it hold usable records it can export?
- Permission: may it license each item?
- Organization: can someone decide, sign and keep records?
Readiness comes before delivery: micro1's Data License Master Terms (updated 2026-11-08) take a worldwide, nonexclusive, irrevocable and perpetual grant, and Labelbox states on its Prism FAQ that "Models already trained on the data cannot be unlearned" (checked 2026-11-09).
Published Entry Thresholds of AI Data Programs
Thirteen programs publish who they accept, and their rules cover headcount, years of records, language, country and company status. The table compares each program's own statement as of 2026-11-09, in alphabetical order, with the role of the publisher.
| Program | Role | Headcount | History | Language and countries | Company status | Source |
|---|---|---|---|---|---|---|
| Collinear AI | buyer program | teams about 50+ | several years of human-written history | English; US, CA, UK, EU, AU, NZ | not stated | collinear.ai |
| Data Licensing Advisors | intermediary (broker) | 30+ desk-based (rejection screen 20+) | 5+ years | US, UK, Canada | not stated; not accounting, legal or healthcare organizations | datalicensingadvisors.com |
| Endless | intermediary (aggregator) | roughly 30+ | several years | English documentation; US, UK, Canada prioritized | not stated | endless-data.com |
| Handshake AI | buyer program | 20+ FTE (license and referral pages); no minimum (Data Partnerships page) | 3+ years | not stated | not stated | joinhandshake.com |
| Labelbox Prism | buyer program | no explicit size rule | not stated | not stated | distressed and wound-down eligible | prism.labelbox.com/faq |
| Marin Data | intermediary (broker) | "most AI labs look for" 20+ | 3+ years, several tools | not stated | winding-down companies may qualify | marindata.ai |
| micro1 | buyer program | 30+ full-time | "substantial historical data" | primarily English; US first, then Western markets | defunct track: 30+ staff, 10+ countries | micro1.ai |
| Mode | buyer program | 30+ (home) vs 50+ (referral page) vs 50 office-based (agreement) | not stated | US strongest; others case by case (home) | not stated | data.mode.inc |
| Nyne | buyer program | 20+ full-time US W-2 | 3+ years of operational workflows | USA, Canada, Europe, Other (selector) | not stated | nyne.ai |
| Pangea.ai | marketplace | mostly 30+ | 2+ years | English; US first | not stated | pangea.ai |
| Pewter | intermediary (broker) | 20+ now or at peak | 3+ years of records | US; mainly English | operating, winding down or closed | pewterdata.com |
| Scale AI | buyer program | 40+ starting point, case by case | not stated | not stated | not stated | scale.com |
| Turing Project Lazarus | reseller to labs (sole seller for 90 days) | not stated | not stated | most regions; strongest demand US, UK, Canada, EU; English sells best | active, winding down, or shut down (90 days or less, or more) | lazarus.turing.com |
Role: brokers place data with buyers and do not acquire it; the aggregator Endless licenses data non-exclusively and places it with several AI labs; "not stated" means the program's page gives no rule.
Headcount, Operating History and Language
The general floors run from 20 to about 50 employees (Mode states 6+ for legal) and from 2 to 5 years of records, and most programs prefer English-language data from the US, UK or Canada. Some programs state no headcount rule: Polyshares states any company with real operating history, Replay states a single requirement, and Defined.ai states no fixed minimum volume (all checked November 9, 2026).
Company Status: Operating, Winding Down or Closed
Several programs accept companies that are winding down or already closed, and some count headcount at the company's peak rather than today. Turing states that Project Lazarus covers wound down companies, and Pewter counts people now or at peak. micro1's referral page states that a defunct company needs 30+ employees, 10+ countries and English-language data (checked November 9, 2026).
Where Published Thresholds Disagree
Three programs state different thresholds on different pages of their own sites, so a company should read the page that governs its route. The statements are shown side by side and not resolved (checked November 9, 2026):
- Mode: the home page states 30+ employees (10+ in accounting, finance and medical; 6+ in legal), the referral page states 50+ (15+ in accounting, medical and finance; 10+ in legal), and the agreement states 50 office-based.
- Handshake AI: the Data Partnerships page states no minimum, while the license and referral pages state 20+ full-time employees and 3+ years.
- Data Licensing Advisors: 30+ desk-based employees, while its rejection screen states 20+.
Data Readiness: Records a Buyer Can Use
Data readiness asks whether the company holds records a buyer wants and can get them out of its systems.
Systems, History and a Record of How Work Was Done
Programs ask for work spread across several systems over several years, because the record of how work was done is what they license. Replay states "Any industry. One requirement": a record of how the work was done, such as communications, tickets and project histories. Scale AI states that it looks for established systems and real workflows (checked November 9, 2026).
SimpleClosure's chief executive states that about 37 % of raw code is removed in cleaning (2026-09-03), a quality signal and not a program rule. What exists, where it lives and how much is the subject of a data inventory for an AI data sale.
Export Access Before Any Approach
A company is not ready until it knows it can export the data on its current plan, because plan limits and vendor terms decide what leaves the system. Four vendor rules show how much the answer varies:
- Slack plans: Free and Pro workspaces export public channels, while private channels and direct messages need Business+ or Enterprise; Free plan file links last 90 days (checked November 9, 2026).
- Slack API Terms: third-party apps "must not… use API Data to train a large language model", and bulk export is prohibited unless an additional agreement allows it (terms of 2025-10-10).
- Google Workspace: the admin account must be at least 30 days old, the export bucket auto-deletes 60 days after the start, and core service data is excluded if the subscription is cancelled or suspended (checked November 9, 2026).
- Microsoft 365 after cancellation: the subscription is expired for 30 days and disabled for 90 days (admins only), and data is deleted no later than 180 days after cancellation (page dated 2026-06-08).
A company that has already cancelled has the least room. Per-system detail sits under export rights and mechanics by source system.
Permission Readiness: Rights, Consent and Personal Data
Permission readiness asks whether the company may license each item, and it removes data from the sale rather than lowering its price.
Ownership and Authority to License
Buyers ask the seller to confirm it owns the data and has not licensed it elsewhere on similar terms, and some ask for contractor IP assignments. micro1's master terms (section 5.1, updated 2026-11-08) have the licensor warrant that it has not granted a third party rights to the same data on the same or substantially similar terms. Patronus AI states owner authority and asks for contractor IP assignments, and Ooak states that the company must own its records (checked November 9, 2026).
Ownership is confirmed item by item in a rights audit.
Customer, Employee and Personal Data
Personal data in customer and staff records is the most common reason a company is not yet ready, because it must be removed or covered by a lawful basis before any license. The CCPA exemption for employee data has been inoperative since 2023-01-01 (Cal. Civ. Code 1798.145(m)(4)). Removing names is not enough: Rocher, Hendrickx and de Montjoye report in Nature Communications 10:3069 (2019-07-23) that 99.98 % of Americans would be correctly re-identified in any dataset using 15 demographic attributes.
The standards are set out under PII removal and de-identification. Labelbox states that medical data requires IRB approval or qualified de-identification (checked November 9, 2026). An FTC Office of Technology staff post (2024-02-13), not a rule, says that retroactive, surreptitious changes to privacy terms to allow AI training may be unfair or deceptive.
micro1's master terms (section 5.7, updated 2026-11-08) exclude these classes unless the Order Form allows them, among other categories:
- Protected health information (PHI)
- Biometric identifiers
- Payment card or financial account data
- Data of children under 18
- Export-controlled information
Data Classes Programs Exclude Before Review
Some programs name data they will not take at all, which settles readiness for those records before any review. Regulr states that it excludes law-firm client files, privileged communications, client tax returns and third-party licensed data, and Data Licensing Advisors does not accept accounting, legal or healthcare organizations (checked November 9, 2026). The wider list is on company data that cannot be licensed for AI training.
Organizational Readiness: Who Decides, Who Signs, Who Keeps the Records
A company is organizationally ready when one person can approve the scope and sign, and the company can meet the duties a license keeps running after delivery. Nyne states that it looks for a clear decision-maker (checked November 9, 2026).
The duties continue after signing. micro1's master terms (updated 2026-11-08) require consent and release records to be kept for 7 years after the agreement ends and security incidents to be notified within 2 business days after the licensor becomes aware. Ooak states that wind-down deals are closed with a director, liquidator or administrator signing.
Consent from the board, shareholders or lenders is a separate step. The route is covered by board and shareholder approval for a data licensing deal, and readiness only asks that the signer is known and the approval route is clear.
The Data Sale Readiness Checklist
Twelve questions cover the four tests, and every No marks work to finish before contacting a buyer.
- Does the company meet the headcount floor of the program it is considering?
- Does its record history match the years that program states?
- Do language, country and company status match the program's page?
- Does it hold records of how work was done across several systems?
- Can each system be exported on its current plan before any deadline?
- Is the data usable after cleaning, without duplicates or third-party content?
- Does the company own the data and hold contractor IP assignments?
- Is it free of earlier licenses on similar terms?
- Are customer, staff and other personal data removed or covered by a lawful basis?
- Are the program's excluded data classes identified and left out?
- Can one named person approve the scope and sign?
- Can the company keep consent records and give incident notice after delivery?
A No is a task and not a failure, and several Nos can mean not selling.
Benefits, Risks, Downsides and Alternatives of Getting Ready, Including Not Selling
Getting ready costs time and money before any offer exists, and the work has value even if the company decides not to license.
Benefits
- An inventory and rights picture useful in any sale or audit.
- Exclusions found before a buyer finds them.
- Comparable approaches to several programs.
Risks
- Sharing samples before rights are clear.
- Treating a program threshold as if it were an offer.
- Exposing personal data during preparation.
Downsides
- Export and de-identification work costs money when no deal follows; no public figure exists.
- A license once delivered cannot be pulled back from trained models.
Alternatives
- Not selling or licensing at all.
- Waiting until rights or consent are settled.
- Licensing a narrower dataset.
- Taking independent advice from an adviser or counsel.
Whether to proceed at all is answered under should you sell or license your data to AI companies.
Where Readiness Sits in the Process to License Company Data
Readiness is step one of the seller process, and its answers feed the inventory, the rights audit and every later step. The full sequence to sell or license company data to AI companies runs through these steps:
- Readiness: this check.
- Inventory and rights audit, with customer consent and confidentiality clauses for customer records.
- Employee notice and consent for staff records.
- How much company data is worth to AI companies.
- Choosing a buyer for your data.
- The data manifest.
- Alternatives to selling data to AI labs, the exit step.
Readiness for a Shut-Down or Closed Company
A closed company faces the same four tests, but export deadlines run from cancellation and the signer may be a director, liquidator or administrator. The route is described under selling data from a shut-down or defunct company. Deadlines are collected under exporting company data before shutting down.
Program Pages Behind the Thresholds
The thresholds above come from these programs' own pages, each checked November 9, 2026.
- AI data partnership payouts and program terms compared
- Company data partnership programs
- Handshake AI data buyer profile
- micro1 data buyer profile
- Who buys AI training data
This page is information, not legal advice.
Frequently Asked Questions
Is there a minimum company size to license data for AI?
No legal minimum exists; programs set their own floors, generally from 20 to about 50 employees, and some state none. Handshake AI's Data Partnerships page states no minimum (checked November 9, 2026). A per-company comparison is the subject of the AI data buyer match.
Does meeting a program's threshold mean a buyer will make an offer?
No; a threshold is an entry rule for an application, and rights, consent and data quality decide whether an offer follows. Scale AI states that its 40+ employees is a starting point, case by case, and no program publishes the share of applicants that receive offers.
How can I tell whether a data licensing approach is legitimate?
Check that the program's terms are published on its own site and that nobody asks for payment or credentials before a contract.
Red flags are collected under is that data licensing email legitimate.