AI training data licensing for businesses

Licence vs outright sale

With a licence you keep ownership and grant a defined right to use a copy of the data, for a stated purpose, under stated conditions. With an outright sale ownership passes to the buyer, who can generally do what they like with it. Licences are the norm for AI training because they let you keep control and, where non-exclusive, license again.

Non-exclusive vs exclusive

A non-exclusive licence lets you offer the same data to others. An exclusive licence gives one buyer sole rights, for a defined scope and period, and should pay more to reflect what you are giving up. Ask how exclusivity is defined: by data source, date range, use or all of them. Exclusivity is one of the factors that affect value.

Key clauses to understand

Scope of data

Exactly which systems, teams, date ranges and file types are included. Anything not listed should be excluded.

Permitted use

What the buyer may do: train models, evaluate them, share with third parties? Narrow and clear is better for you.

De-identification standard

Who removes or masks names, emails and other identifiers, to what standard, and how it is checked before transfer.

Data deletion

When the raw copy must be deleted and how that is confirmed. Note that data already used in training generally cannot be removed from a trained model.

Warranties and indemnities

You will likely be asked to confirm you have the right to license the data. Watch for broad indemnities that leave you liable for issues you cannot control, and look for caps on liability.

Confidentiality

Whether the existence and terms of the deal stay private, and how the buyer protects any commercially sensitive content.

Payment terms

The amount, when it is paid, and any conditions such as approval of the data. For reference, micro1's published range ($100k to $2M+ per data package), subject to micro1's review and approval; not guaranteed.

What to ask your lawyer to check

• That the scope matches what you intend to share, and nothing more
• How permitted use is limited, including onward sharing
• Your warranties, indemnities and liability caps
• Compliance with GDPR, UK GDPR, CCPA or other applicable privacy law
• Conflicts with client, supplier or investor contracts
• Whether staff need to be notified

Platform terms

Your data sits in other companies' software. Some SaaS providers restrict how data exported from, or accessed through, their APIs can be used, and a few specifically address AI training. Check the terms of each tool, such as your chat, ticketing and CRM systems, before including its data.

How micro1 data partnerships fit

micro1 acts between AI labs and companies. It assesses your records, agrees a written scope you approve, de-identifies the data, manages the transfer and pays under the agreed terms once the data is approved. The full sequence is in our guide to selling company data. Advisors who want to introduce clients can read about our partner programme, which also covers companies that have shut down.

Questions to ask before signing a licence

• Scope: which systems, date ranges and record types are included, and is everything else expressly excluded?
• Permitted use: is use limited to training and evaluating models, or are wider uses allowed?
• Exclusivity: is the licence exclusive or non-exclusive, for how long, and does it restrict future deals?
• De-identification: who performs it, to what standard, and can you review samples before transfer?
• Security: how is data transferred, stored and accessed, and who on the buyer's side can see it?
• Retention and deletion: how long may the buyer keep raw data, and what must be deleted at the end?
• Warranties and liability: what are you promising about ownership and rights, and is there a cap on liability?
• Payment: what triggers payment, and what happens if part of the data is rejected in review?

Common mistakes

Businesses sometimes agree to broad warranties about rights without checking contracts with clients, contractors or past acquisitions. Others overlook customer agreements that promise deletion of data at the end of a relationship, which can conflict with licensing. A third mistake is treating exclusivity as a pure upside without considering what it rules out later. Having a lawyer review the draft, especially the warranties, exclusivity and deletion clauses, is usually worthwhile. This guide is general information, not legal advice.

Who to involve internally

A licence is a contract, so it needs the same care as any significant commercial agreement. Bring in whoever owns your key systems, usually IT or an operations lead, so exports and access are not a bottleneck. Involve finance to understand how any proceeds will be treated, and legal counsel or an external advisor to review the scope, warranties and deletion terms. If investors hold consent rights over significant assets, brief them early rather than at the point of signing.

It also helps to nominate one person to coordinate the process. They keep the inventory of systems up to date, collect answers from each team, and act as the single point of contact during scoping. That avoids conflicting answers and makes review faster. Keeping a simple written log of decisions, such as what was excluded and why, gives you a clear record if questions come up later.