Data licensing for AI training

What is data licensing?

Data licensing means giving someone else the right to use your data on defined terms – without giving up ownership. When you license data to an AI company, it is usually about the right to use it as training data for a model.

That is different from selling a physical archive. The data stays with you, you can license it again, and you set the boundaries.

License or sale?

In everyday speech many people talk about "selling data to AI". In practice it is almost always a license, and that is the model we work with.

What a data license covers

A good license is clear on four points.

Exclusivity

We start from non-exclusive licenses. That lets you license the same data to several buyers. Exclusivity can be negotiated, but then comes at a higher price and should be time-limited.

Scope

Which parts of the data are included? A certain number of years, certain channels, certain document types? The clearer the boundaries, the easier it is to assess both value and risk.

Term

The license runs for a defined period. The agreement also covers what happens when it expires.

Use and resale

The license states what the data may be used for, for example training a certain type of model. The buyer may not resell the data or share it with others.

What data can be licensed?

In principle, almost any data you have rights to can be licensed, but the conditions differ.

• Content and media archives – published texts, books, manuals, images, video, audio, source code and structured datasets. Rights are often clear and personal data limited, which makes licensing simpler. Read more about licensing text, image and audio.
• Operational data – email archives, Slack and Teams, documents and intranets, CRM history, support tickets, call recordings and exports from business systems. It is often particularly interesting for training AI agents, but takes more work on personal data and de-identification.

In both cases the same foundation applies: a limited, non-exclusive license where you stay in control.

How the broker model works

CorpusMarket is a broker. We connect European companies that have data with AI companies that license it as training data. We broker – we don't buy your data.

1. You describe your data in a form that takes about two minutes. No files.
2. We assess rights, personal data and demand, and reply within 48 hours on business days.
3. You approve an offer and an agreement. Nothing is shared until a non-disclosure agreement and an accepted offer are in place.
4. Delivery after de-identification, followed by payment.

We get paid through revenue sharing when a deal goes through. It costs nothing to have your data assessed. The whole process is described step by step on how it works.

Rights and personal data first

Before a license is on the table, we go through two questions: do you have the right to license the data, and does it contain personal data? We prioritize data without personal data. Where personal data is present, a well-founded legal basis and de-identification are required – read more about GDPR and anonymization.

For copyrighted content, the EU DSM Directive's text and data mining rules matter: mining is permitted unless the rights holder has reserved those rights. A reservation combined with a license is how free mining becomes paid access.

Documentation buyers need

The EU AI Act (Article 53) requires providers of general-purpose AI models to respect copyright reservations and publish a summary of their training content. Our documentation of origin, rights and terms helps buyers meet those requirements – which makes your data easier to license.

This page is general information, not legal advice. Wondering what your data could be worth? See what is your data worth.