Licensing company data to AI developers in Greece

Short answer

In principle, yes. A Greek company can license data it lawfully holds and controls, in the same way it licenses software or know-how. The hard part is the inventory, not the contract. A typical company archive mixes the company's own records with material clients sent in confidence, personal data, work created by employees and freelancers, bought-in content and open-source code. Each of these has its own owner and its own limits, so before any licence the company has to sort what it can grant freely, what it can grant only after cleaning or consent, and what it must leave out.

Data is not owned like a building

Neither Greek nor EU law gives anyone a single property right over ‘data’. What a company has is a bundle of separate protections, and a data licence is a contract that grants use of that bundle on agreed terms. Copyright may cover texts, drawings, photographs and software. The maker of a database has its own right to prevent extraction of a substantial part of the contents, where it made a substantial investment in obtaining, verifying or presenting them. Information that is secret, has commercial value because it is secret, and has been protected by reasonable steps is a trade secret. And in practice the strongest protection is often the plainest one: the company holds the data, and nobody gets a copy unless the company agrees.

The consequence is simple. A company can license only what it has. If part of the archive belongs to someone else, or is held under someone else's conditions, the licence cannot cure that, and the warranties a buyer will ask for (see our page on licence terms) will expose it.

What is usually yours

Most companies own more than they think, provided the material was generated inside the business and not received from outside it.

Records the company generates in running its business: production and maintenance logs, test and inspection results, engineering calculations, quality files, internal procedures, pricing histories, drawings made in-house.
Databases built at the company's own cost, where the effort went into collecting and checking the contents.
Software written by employees in the course of their employment. Greek copyright law transfers the economic right in a computer program created by an employee under the employment contract, or on the employer's instructions, automatically to the employer, unless the parties agreed otherwise.

Employees, freelancers and the work they leave behind

For works other than software, the rule is narrower and catches many owners by surprise. The employee who writes a report, prepares a drawing or takes a photograph remains its author. Unless the contract says otherwise, only those powers that are necessary to fulfil the purpose of the employment contract pass to the employer. Whether licensing that material to train an AI model falls within the purpose of an engineer's or a technical writer's contract is not obvious, and most Greek employment contracts say nothing about it.

Freelancers and outside consultants are a different case. Nothing passes automatically. The company has what the services contract assigns to it, and if there was no written contract, it may have little more than a right to use the deliverable for the project it was made for. A short review of the employment and service contracts behind the most valuable material is usually the first real task.

What is usually not yours, or not only yours

The largest risk sits in material that came from clients and partners. A shipowner's vessel data, a pharmaceutical partner's study results, a customer's source code, the tender drawings a client sent for pricing: all of these were handed over for a purpose, often under a confidentiality clause or a separate non-disclosure agreement, and many of them are the client's trade secrets. Using or disclosing a trade secret in breach of a confidentiality agreement, or of a duty to limit its use, is unlawful under the trade secrets rules, quite apart from the breach of contract.

Project folders are rarely clean. The company's own calculations sit next to the client's specifications, and the client's data is often embedded in the company's reports. Such files either need the client's written consent or careful separation.

Bought-in content is the next category: purchased standards, subscriptions to technical databases, journal articles, stock images, market reports. The licence under which the company holds them almost never allows the company to pass them on to an AI developer.

Open-source code deserves its own line in the inventory. The company holds it under the licence its authors chose; it cannot grant more than that licence allows, cannot present the code as its own, and some licences attach conditions to any redistribution. A software house's repository usually mixes proprietary code with open-source components, and the two must be identified before anything is offered.

Personal data is a separate question with its own rules, covered on our page on preparing data before licensing.

Public online does not mean free to sell

Since the 2022 amendment of the Greek copyright law, which brought in the EU rules on text and data mining, anyone with lawful access to works may reproduce and extract them for text and data mining, unless the rightholder has expressly reserved that use in an appropriate manner, such as by machine-readable means for content made available online. This exception allows an AI developer to mine. It does not turn content the company collected from the web into the company's property, and it does not allow the company to resell third-party content as a dataset.

The same rule works in the other direction. If the company publishes material online that it may later want to license, a machine-readable reservation of rights helps preserve its value; without one, developers may in principle mine that material without paying.

Sorting the archive before talking to buyers

A short internal exercise, done before the first call with a buyer or an intermediary, avoids most of the problems described above.

Map each data source by origin: generated in-house, received from clients or partners, bought, downloaded, or open-source.
Pull the contracts that govern each source: client agreements and NDAs, employment and freelancer contracts, software and database licences.
Mark the material that is a trade secret of the company itself. Even a sample shown to a buyer can give away more than the price is worth.
Confirm who signs for the company and whether any shareholder or lender consent is needed for a transaction of this kind.

Questions clients ask

Do we need our clients' consent to license data from their projects?

If the data is the client's, or was received under a confidentiality clause, in practice yes, in writing, or the material must be left out. Data the company generated itself while carrying out the project is a closer question that turns on the wording of the client contract.

Our company has been declared bankrupt. Who can sign a data licence?

From the declaration of bankruptcy the debtor loses the management and disposal of its property, which the trustee alone exercises; acts of the debtor without the trustee's participation are ineffective. A data licence is such an act, so it is for the trustee to decide.

Can we license data we collected from public websites?

Rarely. Lawful access may allow text and data mining, subject to rightholders' reservations, but it gives the company no rights it can pass on. Reselling third-party content as a dataset needs the permission of the people who own it.

Sources

This page is general information, not legal advice; the answer in any real case depends on its facts and documents.

L+A advises Greek companies on data protection, data licensing and the regulation of artificial intelligence.

Send us your question

Tell us about your case. A lawyer of the firm will reply by email.

Your message goes only to the lawyers of the firm and is used only to reply to you. You may write in Greek or English.

Can my company license its data to AI developers, and how much of that data is actually ours to license?