# Text and Data Mining (TDM)

> A computer copies lots of text to find patterns. EU law allows it, with limits.

**Field:** Artificial Intelligence  
**Category:** Ethics & Regulation  
**Also known as:** TDM, TDM Exception, Text and Data Mining Exception

## Think of it like

It is like reading all the books in a room. You count which words come up most.

## Definition

Text and data mining is when a computer reads huge amounts of text or data to find patterns. In EU copyright law, it is also a rule that lets you copy for this without asking. Research bodies may mine any work they can lawfully reach, for research. Anyone else may too, unless the owner has said no in a way a machine can read. AI makers lean on it to train models on the web. The EU AI Act says model makers must respect those refusals.

## Example

A school in Amsterdam mines 2 million news stories to see how words change over time. As a research body, it needs no one's okay.

## Why it matters

It is the main legal door AI makers use to train on public web pages in the EU. The owner's opt-out is what closes it.

## Common mistake

Thinking the exception works the same everywhere. It is EU law, other countries have their own rules, and an opt-out can switch it off.

## Three principles

- **Find patterns.** Computers read at a huge scale.
- **Two exceptions.** One for research, one for anyone.
- **Opt-out wins.** Outside research, owners can say no.

## Related terms

- [AI Training Opt-Out](https://learn.benibix.com/term/ai/ai-training-opt-out)
- [Training Data Copyright](https://learn.benibix.com/term/ai/training-data-copyright)
- [DSM Copyright Directive](https://learn.benibix.com/term/business/dsm-copyright-directive)
- [Fair Use](https://learn.benibix.com/term/business/fair-use)
- [Web Scraping](https://learn.benibix.com/term/development/web-scraping)
- [General-Purpose AI Model (GPAI)](https://learn.benibix.com/term/ai/general-purpose-ai-model)

## Told apart from

- [Fair Use vs Text and Data Mining Exception](https://learn.benibix.com/compare/fair-use-vs-text-and-data-mining): People think the EU's data mining rule is Europe's version of fair use, and it is not. Fair use is a US rule where a court weighs four factors, case by case. The EU rule is a fixed exception for copying to find patterns. Anyone may use it, unless the owner opts out in a form a machine can read. A fair use win in the US says nothing about the EU.

![Two questions in a column. First: a research body, mining for research? Yes leads to allowed, and the owner cannot opt out. No leads to the second question: did the owner say no in a form a machine reads? No leads to allowed, anyone may mine it. Yes leads to ask first, you need the owner's licence. A note says every path starts from lawful access to the work.](https://learn.benibix.com/diagrams/ai4-text-and-data-mining.svg)

---

Source: https://learn.benibix.com/term/ai/text-and-data-mining
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-18
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
