AI Glossary

All 1 A B C D E F G H I J K L M N O P Q R S T U V W Z

T

Training Data

Data used to teach an AI model during its development.

Training Data Summary Template

A template published by the EU AI Office according to which providers of GPAI models disclose a sufficiently detailed overview of the main sources of training data, so that rights holders can exercise their rights.

Training Dataset

A dataset used to train AI models.

Training Scraping

The automated collection of content from the internet for the purpose of training models; it raises questions of compliance with copyright law.

Transfer of Personal Data to a Third Country

The provision of personal data outside the European Economic Area.

Transformative Use

A fair use factor assessing whether a new use adds a different purpose or character compared to the original work. In the Bartz case, training a model was deemed highly transformative.

Transformer

A type of neural architecture on which modern language models are built. It processes input data in parallel and uses an attention mechanism to determine the relationships between elements of the text.

Transitional Provisions of the AI Act

Rules determining how and from when obligations apply to systems and models already placed on the market before the individual parts of the regulation take effect.

Transparency Chapter (GPAI Code)

A part of the code applying to all providers of GPAI models that clarifies how to meet the information obligations under Article 53, in particular by providing documentation to downstream providers and the AI Office.

Trial Operation of AI

Verification of an AI system in a limited and controlled environment before wider deployment.

Trustworthy AI

A concept of AI that is lawful, ethical, and technically robust; it includes, among other things, respect for fundamental rights, transparency, fairness, and human oversight.

Trustworthy AI Ecosystem

An environment in which rules, oversight and user trust support the uptake of responsibly developed and deployed AI; a goal toward which European regulation is directed.

Trustworthy AI under the OECD Principles

An intergovernmental set of value-based principles for trustworthy AI that has become a reference point for many national and supranational frameworks and unifies the basic terminology.

Turnover-Based Fine

A method of calculating a penalty in which the maximum fine is based on a company's total worldwide annual turnover, so that the sanction remains effective even against large companies.