Technical

text-and-data mining

Automated analysis of text and datasets to extract information.

How the laws define this

Text-and-data mining is uniformly recognized across these frameworks as the automated techniques or processes used to analyze large collections of digital text and data to extract or generate information, patterns, trends, or insights, inherently raising questions around copyright and intellectual property rules. The primary divergence lies in how different legal instruments handle copyright exceptions and permissible use cases. Law No. 2016-1321 of 7 October 2016 - Law for a Digital Republic (Loi pour une République numérique) restricts its statutory exception strictly to non-commercial scientific research under defined conditions. In contrast, the statement under Section 137 Data (Use and Access) Act addresses the broader debate surrounding potential copyright exceptions for commercial AI development and mechanisms for rights holders to opt out. Meanwhile, the Information Session connects text-and-data mining for model training, statistical analysis, and indexing to obligations under the DSM Directive and AI Act, whereas the WIPO Conversation on AI and Intellectual Property Policy focuses on licensing and copyright exceptions for AI training data generally.

Synthesised from the 4 statutory definitions below. Each one is quoted in full, with its source.

How jurisdictions differ

What is distinctive in each definition, binding law in force first. Showing 3 of the 4 instruments that define this term; all are listed under Definitions below.

Jurisdiction	Instrument	Status	What is distinctive
France	Law No. 2016-1321 of 7 October 2016 - Law for a Digital Republic	In Force (Amended)	Adds a limited exception permitting TDM for non-commercial scientific research under defined conditions, clarifying its lawful scope against intellectual property rules.
European Union	Information Session: Protocols for Text and Data Mining Rights under AI Act and GPAI Code of Practice	Draft	Covers activities implicating copyright reservations, with protocols for opt-outs described under the DSM Directive and AI Act on training data.
United Kingdom	Copyright and artificial intelligence statement of progress under Section 137 Data (Use and Access) Act	Under Review	Emphasizes TDM's role in AI model training with copyrighted material, addressing debates on commercial exceptions and rights holder opt-out mechanisms.

What it triggers

In France, under Law No. 2016-1321, it permits text and data mining for non-commercial scientific research under defined conditions.source

Definitions (4)

Automated methods and processes for analyzing large corpora of text or data to extract patterns, information or insights; the Act establishes a limited exception permitting TDM for non‑commercial scientific research under defined conditions. The definition clarifies the lawful scope of automated analytical activities vis‑à‑vis intellectual property rules.

Law No. 2016-1321 of 7 October 2016 - Law for a Digital Republic (Loi pour une République numérique)Definition 1 of 4

Automated techniques and processes used to analyze, extract, or infer information from large collections of text or data for purposes such as indexing, statistical analysis, or model training. In this context, TDM refers to activities that may implicate copyright reservations and are covered by the DSM Directive and the AI Act obligations on training data.

Information Session: Protocols for Text and Data Mining Rights under AI Act and GPAI Code of PracticeDefinition 2 of 4

Text and Data Mining (TDM) is a crucial concept, referring to the automated analytical techniques used to analyze text and data in digital form to generate information which includes patterns, trends, and correlations. For AI development, TDM involves using vast datasets, often containing copyrighted material, to train AI models. The statement addresses the debate around a potential copyright exception for TDM, particularly for commercial purposes, and the mechanisms for rights holders to 'opt-out' of such use.

Refers to the automated analytical techniques aimed at analyzing text and data for patterns, trends, and other information, which is highly relevant for AI training data and raises questions about copyright exceptions and licensing.

WIPO Conversation on AI and Intellectual Property PolicyDefinition 4 of 4

Not to be confused with

Data analysis

Data analysis refers to statistical, predictive and pattern-recognition methods applied to data, while text-and-data mining specifically involves automated analysis of large text corpora or datasets.

research data

Research data is data produced by research, especially publicly funded, whereas text-and-data mining analyzes such data using automated techniques.

Questions and answers

What exception does French law provide for text and data mining?

Law No. 2016-1321 establishes a limited exception permitting TDM for non-commercial scientific research under defined conditions. This clarifies the lawful scope of such automated analytical activities vis-à-vis intellectual property rules.

How does the EU address TDM rights under the AI Act?

It describes TDM activities that may implicate copyright reservations through technical protocols for opt-outs, under the DSM Directive and AI Act on training data.

What is the UK position on TDM for AI training?

TDM involves using vast datasets with copyrighted material to train AI models, with the statement addressing debates on commercial copyright exceptions and rights holder opt-out mechanisms.