Sell Data to AI
Home Data Asset Score Pricing API documentation US labs and CROs list
For brokers
How to become an AI data broker Data broker business model Buyer programs compared Qualify a company AI training data companies
For data companies
Firmographic data providers Company data API
Seller guides
How to sell data to AI companies Is it legal? FAQ and glossary About
Check domain/company
Contracts, clause by clause

The AI Data Licensing Agreement, Clause by Clause

Last checked: 7 October 2026 (buyer statements quoted on this page)

This page outlines what a data licensing agreement with an AI buyer usually covers, and the questions to ask under each heading. It is not a template and there is no text to copy: your own lawyer should draft or review the actual words.

14clause groups outlined, each with the questions to ask
9contract risks every seller should check, whoever the buyer is
60 to 90 daystypical close cited by practitioners, for the whole deal
Where the agreement sits

The agreement is step four of eight

Practitioners describe the same sequence for most company data deals, and cite 60 to 90 days from start to close. Nobody can promise speed.

Inquiry
NDA
Buyer review
Agreement
Export
De-identify
Acceptance
Payment

Everything after step four runs on the terms you sign at step four.

Before the agreement, the buyer usually wants to see what you have. Practitioners give one firm rule here: never send a full dataset before price. Share a manifest (a list of systems, date ranges and volumes) and small samples, and get more than one offer. Comparing two agreements side by side is the fastest way to see which terms are negotiable; see getting more than one offer.

The paperwork may come as one document or several. Whatever the format, the topics below need answers somewhere in it.

NDA Main license or master agreement Data schedule or order form Data protection terms Acceptance criteria
No template here

Why this page does not offer a template

People search for "data licensing agreement template" more than for anything else on this topic. Here is why we outline instead.

Deals differ where it matters

Scope, permitted use, exclusivity and liability change from deal to deal. Those are exactly the points a generic form glosses over.

Forms give false comfort

Templates written for licensing photos, software or marketing lists do not ask who pays when de-identification misses a name in a chat thread.

Your facts decide the words

Your client contracts, staff notices, industry rules and location shape every clause. Only your own lawyer can turn those facts into text.

The outline

Fourteen clause groups and the questions to ask under each

Names and order vary between contracts. Use this as a checklist: if a topic is missing from your draft, ask where it is handled.

Part A · What is licensed

Clause 01

Parties and definitions

Covers: Who licenses and who receives, including any affiliates, and the defined terms the rest of the contract depends on, such as "licensed data", "derivatives" and "models".

  • Does the receiving side include affiliates or unnamed partners?
  • Does the definition of the licensed data match your scope list exactly?
  • Is "derivative" defined, and does it include trained models?
Clause 02

Data scope and schedule

Covers: The systems, teams, date ranges, file types and volumes in the deal, and what is excluded. This often sits in an attachment rather than the main text.

  • Is there a written exclusion list (HR, legal, client-confidential, private messages)?
  • Who on your side signs off the export?
  • Can scope only grow by a signed amendment?
Clause 03

License grant and permitted use

Covers: What the buyer may do with the copy: train, fine-tune, evaluate, benchmark or build products on it, and for how long.

  • Is use limited to training, or does it include evaluation and commercial products?
  • Are some uses excluded, such as identifying individuals?
  • Does the grant run for a fixed term or without end?
Clause 04

Exclusivity, resale and onward transfer

Covers: Whether you may license the same data to others, and whether the buyer may pass it on, for example to AI labs it supplies.

  • None, time-limited or perpetual exclusivity?
  • Does exclusivity cover only this copy, or similar data you create later?
  • Who receives the data downstream, and are they bound by the same limits?
Read more: Exclusivity and resale rights
Clause 05

Ownership and intellectual property

Covers: Who owns the original records, the processed copy, any labels or annotations added to it, and what is built from it.

  • Does the contract itself state that you keep ownership of the original records?
  • Who owns the de-identified copy?
  • Do you keep any rights in what the buyer builds?

Part B · Money and delivery

Clause 06

Payment

Covers: Price, structure and timing: one-off or recurring, milestones, currency, taxes and payment method.

  • Is money due at signing, at delivery or at acceptance?
  • What triggers each milestone payment?
  • If part of the data is rejected, how is the price adjusted, and are there refund or clawback terms?
Clause 07

Delivery and acceptance

Covers: How data is exported and handed over, what quality checks the buyer runs, and how long it has to accept or reject.

  • Are acceptance criteria written down and measurable?
  • How long is the review window, and what happens if it passes in silence?
  • Is rejected data returned, deleted or kept?

Part C · Privacy and risk

Clause 08

De-identification and audit rights

Covers: Who removes personal and confidential details, by what method, at which step, and how you can check the result.

  • Who performs it, and is a third party involved?
  • Can you review a de-identified sample of your own data before release?
  • Can you get a written method description, a report or an audit right?
Read more: De-identification before selling
Clause 09

Confidentiality and publicity

Covers: What each side must keep secret, including the existence of the deal, and whether the buyer may name you.

  • Does the clause respect confidentiality you owe your own clients?
  • May the buyer announce the deal or use your name?
  • How long do confidentiality duties last?
Clause 10

Representations and warranties

Covers: Statements of fact you promise are true: ownership, your right to license, notice to or consent from employees and customers, no breach of other contracts, compliance with law.

  • Are your promises limited to what you know after reasonable checks?
  • Do they still apply to data the buyer has changed?
  • What does the buyer promise in return about security and use?
Clause 11

Indemnities and limitation of liability

Covers: Who pays for losses from defined events, such as a privacy claim, and the most each side can owe.

  • Who pays if de-identification misses something?
  • Is your liability capped, and how does the cap compare with the price?
  • Is any category uncapped, and is the indemnity mutual?
Read more: Indemnities and warranties
Clause 12

Security, retention and deletion

Covers: How the copy is stored and protected, how long it is kept, and the deletion of originals and copies.

  • Which copies are deleted, when, and with what written confirmation?
  • Does deletion reach downstream recipients?
  • What happens to models already trained on the data, and is there a breach-notice duty?

Part D · The exit

Clause 13

Term, termination and survival

Covers: How long the agreement runs, how either side can end it, and which clauses survive after it ends.

  • Can you terminate if the buyer breaches the use limits?
  • Which obligations survive, and for how long?
  • Does the license to delivered data continue after termination?
Clause 14

Governing law and disputes

Covers: Which law governs the contract, where disputes are heard, and whether arbitration applies.

  • Is the chosen law and venue practical for a company your size?
  • Does your lawyer practice under that law, or do you need local counsel?
  • Is there a negotiation or mediation step before formal proceedings?
Check the text

A website statement is not a clause

Buyers publish privacy statements on their own pages. For any buyer, the useful step is to find where each statement you rely on appears in the agreement you sign.

What buyers publish

  • micro1: scope agreed in writing; sensitive and confidential information scrubbed; originals deleted after processing; no customer information exposed; the company keeps ownership of its underlying data.
  • Mode: buys "an agreed copy"; originals stay with the company; de-identifies before onward delivery.

As published on each buyer's own page, checked 7 October 2026.

Where to look in your draft

  • Scope in writing: the data schedule (clause 2).
  • Scrubbing and de-identification: clause 8, plus your right to review a sample.
  • Deletion of originals and copies: clause 12, with written confirmation.
  • Ownership of underlying data: clause 5.
  • Onward delivery: clause 4, including who the downstream recipients are.

If a point that matters to you is not in the draft, ask for it in writing. Asking is normal; whether a buyer agrees is its decision, and you can compare that answer with other offers.

Before you negotiate

Have these ready before the first draft arrives

Each item answers a question the agreement will ask. Having them early shortens review with your own lawyer.

  • A manifest: systems, date ranges, volumes, owners. No raw data.
  • An exclusion list: HR, legal, client-confidential and regulated records.
  • Client contract review: NDAs and service terms that limit secondary use.
  • A staff notice decision: what employees will be told, and when.
  • Your limits: exclusivity you would accept, and the liability cap you need.
  • Internal sign-off: who may sign, and whether the board or partners must approve.
  • A sample plan: what small, de-identified sample you will share under NDA.
  • Your buyer questions: see questions to ask a data buyer.
Reading a draft

Red flags and green flags in a data licensing draft

A red flag is a reason to ask why and negotiate, not an automatic reason to walk away. A green flag is a term that makes the deal easier to defend later. Both columns describe generic patterns, not any buyer's terms.

TopicRed flag: ask whyGreen flag
Scope"All company data", or no schedule attached at signingA dated schedule listing systems, date ranges and exclusions
Permitted useAny purpose, with no list of uses or exclusionsNamed uses, such as training and evaluation, and named exclusions
ExclusivityNo end date, covering "similar" data you create laterNone, or time-limited and tied to the delivered copy
Onward transferUnrestricted resale to recipients you never learn aboutRecipients bound by the same limits; de-identified copy only
PaymentEverything paid after an open-ended acceptance reviewWritten acceptance criteria and a fixed review window
De-identificationMethod not described; no chance to see outputMethod in writing; you review a sample of your own data
LiabilitySeller-only indemnity with no capMutual indemnities and a cap you can state as a number
DeletionNo date and no confirmationDated deletion of originals, copies and keys, confirmed in writing
SurvivalWarranties with no end dateA stated survival period for each obligation

Patterns compiled from the generic contract risks every seller should check. They are not claims about any named company.

Common mistakes

Six mistakes sellers make before and during the agreement

None of these needs a lawyer to avoid. All of them are harder to fix after signing.

Skimming the NDA

An NDA can carry more than confidentiality, such as an exclusivity or no-shop period during talks. Read it with the same care as the main agreement.

Sending everything to get a price

Practitioners advise a manifest and samples first, and more than one offer. A full dataset sent early is hard to take back.

Trusting the website, not the text

Privacy statements on any buyer's site help only if the signed agreement reflects them. Find each one in the draft.

Leaving the schedule for later

If scope is decided after signing, the grant may already cover more than you meant. Attach the schedule before you sign.

Ignoring acceptance terms

Payment tied to acceptance without written criteria can delay or reduce what you receive, with no clear way to dispute it.

Letting one person decide

A data deal touches legal, IT, HR and client relationships. Agree internal sign-off before the first draft arrives.

Eight questions for your lawyer when the first draft arrives

Send the draft, your scope schedule, your exclusion list and these questions together. Written answers are easier to compare across offers.

  • Does the license grant match our scope schedule exactly?
  • What does "derivative" cover, and does it reach trained models?
  • Is any exclusivity limited both in time and in the data it covers?
  • Which of our warranties can be limited to what we know?
  • What is our maximum exposure, including carve-outs from the cap?
  • Which clauses survive termination, and for how long?
  • Does anything conflict with our client contracts or staff notices?
  • Is the governing law workable, or do we need local counsel?
FAQ

Questions about AI data licensing agreements

Is there a template for an AI data licensing agreement?

Not on this site, on purpose. These deals differ in scope, permitted use, exclusivity and liability, and a generic template can give false comfort on exactly the points that matter. Use the clause outline on this page to prepare your questions, and have your own lawyer draft or review the actual text.

Is a data deal a sale or a license?

The buyers’ own pages describe a copy, not a transfer: Mode states it buys “an agreed copy” and that originals stay with the company, and micro1 states that the company keeps ownership of its underlying data (as published, checked 7 October 2026). Your signed contract decides what you actually grant, so read the grant and ownership clauses closely.

Which clauses matter most for a seller?

The questions every seller should check cluster in five places: permitted use, exclusivity and resale, warranties about consent, indemnities and liability caps, and deletion of originals and copies. Payment and acceptance terms then decide when, and whether, you are paid in full.

Who drafts the agreement?

Either side can. If the buyer sends its standard form, you can still ask for changes; whether it agrees is up to the buyer. Ask your own lawyer to review the points in this outline before you sign anything.

How long does negotiating the agreement take?

Practitioners cite 60 to 90 days to close a whole deal, from NDA and buyer review through the agreement, export, de-identification and acceptance. The agreement is one stage of that. Nobody can promise a timeline.

Should I send the full dataset before the agreement is signed?

Practitioners advise against it: share a manifest and samples, and get more than one offer before any full dataset changes hands. Keep any sample you share small, de-identified and covered by an NDA.

Does signing an NDA commit us to the deal?

An NDA covers confidentiality during talks; it is not the license itself. Some NDAs and letters of intent also add terms such as an exclusivity or no-shop period during negotiation, so read them before you sign. Ask your lawyer whether anything in them binds you beyond confidentiality.

Can we limit which AI companies or models use our data?

You can ask. Permitted-use and onward-transfer clauses can name allowed uses, excluded uses and the types of recipient allowed. Whether a buyer agrees depends on its business model, so compare the answers from more than one buyer before you choose.

See a real draft before you decide

The only way to read a buyer's actual terms is to apply and reach the agreement stage. Check your fit first, then apply to more than one program so you can compare drafts. Applying commits you to nothing.

Independent site. Some links are referral links: if your company signs with a buyer through them, the buyer may pay us a fee. You are not charged, and we never see your data.

Keep reading

Deeper guides to the hardest clauses