Registries for Training Preferences

A Simple, Standards-Based Approach

Creators and rightsholders can make their AI training preferences heard using a registry-based model for rights declarations. At the heart of this model lies a simple yet powerful principle: declare your intent once, and make it globally discoverable.

This is made possible through federated content registries – public, open, and verifiable directories that store and disseminate metadata about how digital works may be used, particularly in contexts like text and data mining (TDM) and AI model training.

TDM·AI supports declarations built on a shared vocabulary and identifier system, including:

• ISCC codes for uniquely identifying content

• Standardized opt-out expressions

• Machine-readable formats for easy integration and compliance

Why This Matters

In today's digital ecosystem, works are collected, processed, and sometimes reused without direct contact between the creator and the user. Registries provide a missing link: a reliable way to declare rights and usage restrictions in a standardized and resolvable format.

This approach offers:

• Transparency: Declarations are publicly visible and verifiable.

• Consistency: A common format ensures uniform interpretation across platforms.

• Scalability: Billions of records can be indexed and queried efficiently.

• Neutrality: No dependency on any specific vendor, platform, or technology provider.

With a registry, the declaration lives outside any individual website or platform, increasing the likelihood of recognition and enforcement.