TDM·AI Protocol

THIS IS A TEMPORARY DRAFT FOR DISCUSSION Last updated 2024-06-17

Summary

TDM·AI is a protocol to inextricably bind machine-readable opt-out declarations for text and data mining (TDM) for the purpose of training models and applications of generative AI to the digital media content. It is based on the DSM Directive on Copyright 2019/790, Article 4, making use of the benefits of the International Standard Content Code (ISCC) and Creator Credentials.

Abstract

The TDM·AI protocol aims to provide rightsholders with a simple and standardised way to make a machine-readable declaration as to whether or not their content may or may not be used for AI training purposes. This opt-out declaration – or its revocation – can be derived directly from the content, which means that it is easily accessible for all AI providers using open source identifier technology.

This protocol utilises the International Standard Content Code (ISCC), a published ISO standard for the content-derived identification of digital media content (ISO 24138:2024) and verifiable Creator Credentials, based on W3C recommendation for cryptographically verifiable credentials, to ensure verifiable and machine-readable declarations that include proper attribution of claims. The protocol also supports time-stamping of this statement.

By using the ISCC, the protocol ensures a decentralised and reliable method of identifying content, making it impervious to common problems such as metadata loss or content modification. The integration of verifiable creator credentials adds another layer of trust and verifiability, ensuring that the declarations are genuine and can be traced back to the original rights holder.

Motivation

Given the evolving AI landscape, there is an urgent need for an protocol that allows content creators and rightsholders to reliably declare their consent or reservation to TDM for AI providers. The TDM·AI protocol is motivated by the need to:

• Provide clarity and simplicity for rightsholders to declare their TDM reservation with regards to AI training,

• Offer a decentralised, immutable, and verifiable system for these declarations,

• Ensure that AI providers and other stakeholders can easily understand and respect these declarations.