Text and data mining ("TDM")

Text and data mining, or fouille de textes et de données, means the use of an automated analytical technique applied to texts and data in digital form in order to generate information. Articles 3 and 4 of Directive (EU) 2019/790 make it the subject of two mandatory exceptions to copyright and to the sui generis right in databases, transposed in France in Article L. 122-5-3 of the French Intellectual Property Code and, for databases, in Article L. 342-3 of the same Code.

The two exceptions

The first exception benefits research organisations and cultural heritage institutions, for reproductions made for the purposes of scientific research from works to which they have lawful access. It is mandatory, rightholders being unable to opt out, and the copies may be stored securely for the purpose of verifying research results. The second exception is open to any person, whatever the purpose of the mining, including a commercial one, for works to which that person has lawful access, but it applies only if the rightholder has not opted out. The copies made may be retained only for as long as is necessary for the mining.

The rightholder’s opt-out

The opt-out from mining for general purposes must be expressed in an appropriate manner. For content made available online, the EU text refers to machine-readable means, such as metadata or the terms and conditions of a website, and practice has moved towards exclusion protocols and declarations in access conditions. The French text refers to an opt-out expressed by any means, including machine-readable means. Lawful access remains a precondition: circumventing an access restriction or a technological protection measure cannot found the benefit of the exception.

The artificial intelligence issue

The training of generative artificial intelligence models relies on the mass reproduction of content, which developers bring under the general mining exception. Regulation (EU) 2024/1689 on artificial intelligence requires providers of general-purpose models to put in place a policy to comply with copyright law, in particular to identify and respect the opt-outs expressed, and to publish a sufficiently detailed summary of the training data. Whether the exception covers all training operations, and whether the model’s outputs may constitute reproductions, remains debated before the courts and among scholars.

Key points

Text and data mining is free for the scientific research of eligible organisations and, for any other purpose, in the absence of an opt-out by the rightholder, subject to lawful access. Rightholders have an interest in expressing their opt-out by machine-readable means and in their terms of use. Operators of artificial intelligence systems must document their sources and check for opt-outs before any training.

---

This glossary entry is general in scope and does not constitute legal advice.

Dhenne Avocats is a patent litigation firm. Matters outside that field, such as trade marks, designs, copyright and IT contracts, are handled by practitioners the firm works with regularly, and by the firm itself when they arise in a patent dispute.

Intellectual property attorney · Tell us about your situation