ODRL Profile: AI Vocabulary

This ODRL Profile defines AI-specific actions which enables policy authors to clearly define acceptable and prohibited uses of assets in AI workflows. It supports governance, risk management, and compliance efforts, helping ensure that AI systems operate within agreed legal, ethical, and contractual boundaries.

Prefix	Namespace	Definition
odrl	http://www.w3.org/ns/odrl/2/	[[!odrl-model]] [[!odrl-vocab]]

Current AI Vocabularies

AI "Opt-out" Vocabularies

Numerous standards groups and other projects have developed vocabularies for "opt-out" of AI processes over user content. An analysis has found them to be divergent, fragmented, and imbalanced and also provides a detailed summary of these vocabulary options.

Below is a summary of the key proposals.

IETF - AI Preferences [[?draft-ietf-aipref-vocab]]

AI Training (train-ai) - Using an asset to modify the learned parameters of a generative AI model
AI Use (ai-use) - Using an asset as input to a generative AI model, where the asset is not directly provided by the user model
Search (search) - Using an asset in an application where the primary purpose of the application is to select assets and direct users to the location of those assets

Creator Assertions Working Group (CAWG) - Training and Data Mining Assertion [[?cwag-train]]

cawg.data_mining - Can any text or data content be extracted from the asset for purposes of determining “patterns, trends, and correlations.”
cawg.ai_inference - Can the asset be used as input to a trained AI/ML model for the purposes of inferring a result
cawg.ai_generative_training - Can the asset be used as training data to an AI/ML model that could generate assets.
cawg.ai_training - Can the asset be used as data to train non-generative AI/ML models, such as those used for classification, object detection, etc.

PLUS License Data Format - Data Mining [[?plus-data-mining]]

DMI-ALLOWED - Data Mining allowed
DMI-PROHIBITED - Data Mining prohibited
DMI-PROHIBITED-AIMLTRAINING - Prohibited for AI/ML training
DMI-PROHIBITED-GENAIMLTRAINING - Prohibited for Generative AI/ML training
DMI-PROHIBITED-EXCEPTSEARCHENGINEINDEXING - Prohibited except for search engine indexing

Really Simple Licensing (RSL) - AI Usage [[?rsl-ai]]

all - Any automated processing, including AI training and search
ai-all - Any use by AI systems, including training, inference, indexing, retrieval, generation, grounding, or model evaluation. This term explicitly includes the ai-train, ai-input, and ai-index usage types defined below, as well as other AI-related uses not yet enumerated
ai-train - Training or fine-tuning AI models
ai-input - Input into AI models, including retrieval-augmented generation, grounding, or other use of the content to produce generative AI responses or search summaries
ai-index - Inclusion in an AI system’s internal index or retrieval database
search - Building a search index and providing search results (e.g., returning hyperlinks and short excerpts from your website's contents). Search does not include providing AI-generated search summaries

NOTE: Search and Index in the above systems typically are general concepts and not specific to "AI systems".

TDM Reservation

The TDM Reservation Protocol [[?TDM-Res]] defines a simple protocol capable of expressing the reservation of rights relative to text & data mining (TDM) applied to Web content, and to ease the discovery of TDM licensing policies associated with such content. The specification already uses the W3C ODRL 2.2 Policy Language, and only explicitly defines this vocabulary term:

Text & Data Mine (mine) - analyse, via automated analytical technique, text and data in digital form in order to generate information which includes but is not limited to patterns, trends and correlations

ODRL AI Vocabulary Profile

As can be clearly seen by these existing approaches in Section 3, they vary in consistency in their model and semantics, and hence will dilute data interoperability for AI governance across the Web and Enterprise.

Using the current set of AI vocabularies described in Section 3 as a snapshot of the AI community requirements, the ODRL AI Vocabulary Profile takes a more bigger-picture view by providing an extensible Data Processing framework that captures current state needs, as well as sets the platform for future growth. The figure below shows the framework and current scope of the AI vocabulary, including their hierarchy in the framework.

The current scope of the AI Vocabulary consists of six new terms and includes two existing ODRL Vocabualry terms. The new terms also represent a hierarchy, enabling ODRL policies to more simply include the inherited terms. The terms are summarised here:

Artificial Intelligence - the asset is used in Artificial Intelligence computational systems

AI Training - the asset is used to train/modify an AI model

Generative AI Training - the asset is used to train/modify a Generative AI model

AI Usage - the asset is used to perform tasks with AI systems

AI Augmentation - the asset is used as input to a trained AI model

Data Mining - the asset is used for algorithmic analysis
Index - the asset is registered in a data store
Search - the asset is discoverable in a data store

NOTE: Data Mining is not considered as the parent term in this framework, as that is not the industry consensus.

The next sections formally defines these vocabulary terms.

Artificial Intelligence (AI)

Definition:	The act of using asset(s) with computational systems capable of performing artificial intelligence tasks
Label:	Artificial Intelligence (AI)
Identifier:	http://www.w3.org/ns/odrl/2/ai
Note:	Artificial Intelligence encompasses a broad range of techniques and systems, including machine learning, deep learning, and generative AI systems.
Included In:	Use
Class:	Action
Mapping:	This vocabulary term has similar semantics to: RSL: ai-all

AI Training

Definition:	The act of using asset(s) to modify the parameters, internal representations, or learned behaviour of an artificial intelligence model or system to enable or improve its performance
Label:	AI Training
Identifier:	http://www.w3.org/ns/odrl/2/ai-training
Note:	AI Training includes initial model training, continued pre-training, fine-tuning, and other training processes that modify a model based on data
Included In:	Artificial Intelligence (AI)
Class:	Action
Mapping:	This vocabulary term has similar semantics to: CAWG: cawg.ai_training IETF: train-ai RSL: ai-train

Generative AI Training

Definition:	The act of using asset(s) to train or modify a generative artificial intelligence model so that it can generate new content, including text, images, audio, video, software code, or other forms of output
Label:	Generative AI Training
Identifier:	http://www.w3.org/ns/odrl/2/ai-train-gen
Note:	Generative AI Training includes initial training, continued pre-training, fine-tuning, and other training processes that modify a generative AI model
Included In:	AI Training
Class:	Action
Mapping:	This vocabulary term has similar semantics to: CAWG: cawg.ai_generative_training

AI Usage

Definition:	The act of using asset(s) to an artificial intelligence system to perform a task or obtain an output, without necessarily modifying the system’s learned model parameters
Label:	AI Usage
Identifier:	http://www.w3.org/ns/odrl/2/ai-usage
Note:	AI Usage includes activities such as prediction, classification, summarisation, question answering, content generation, and decision support by providing information to an AI system
Included In:	Artificial Intelligence (AI)
Class:	Action
Mapping:	This vocabulary term has similar semantics to: IETF: ai-use

AI Augmentation

Definition:	The act of using asset(s) as input to pre-trained AI models to assist, enhance, or extend their capabilities in performing a task, while retaining involvement in evaluating, directing, or acting upon the resulting output
Label:	ai-augment
Identifier:	http://www.w3.org/ns/odrl/2/ai-augment
Note:	AI Augmentation includes activities in which AI supports human or computational capabilities, such as assisting with research, improving decision-making, generating recommendations, supporting software development, or enhancing creative work. NOTE: This term is sometimes referred to as inference.
Included In:	AI Usage
Class:	Action
Mapping:	This vocabulary term has similar semantics to: CAWG: cawg.ai_inference RSL: ai-input

Data Mining

Definition:	The act of using asset(s) in analysing or computationally processing data to identify, extract, or derive patterns, relationships, trends, correlations, structures, or other information that may not be explicitly represented in the original data
Label:	data-mining
Identifier:	http://www.w3.org/ns/odrl/2/data-mining
Note:	Data Mining may involve statistical analysis, pattern recognition, clustering, association analysis, or other computational techniques
Included In:	Use
Class:	Action
Mapping:	This vocabulary term has similar semantics to: CAWG: cawg.data_mining PLUS: DMI-ALLOWED TDM: mine

Search

Definition:	The act of enabling the asset(s) to be discoverable in a data store
Label:	Search
Identifier:	http://www.w3.org/ns/odrl/2/search
Note:
Included In:	Use
Class:	Action
Mapping:	This vocabulary term has similar semantics to: IETF: search RSL: search

Index

Definition:	The act of enabling the asset(s) to be recorded in the index of a data store
Label:	Index
Identifier:	http://www.w3.org/ns/odrl/2/index
Note:	For example, to include a link to the Asset in a search engine database after the Asset has been processed by an algorithm. This term is the same as defined in the ODRL Common Vocabulary.
Included In:	Use
Class:	Action
Mapping:	This vocabulary term has similar semantics to: RSL: ai-index

Use Cases

AI Examples

The asset (dataset1) maybe used for AI Training only (which will include Gen AI Training) for the explicit purpose of "Research and Development"

{
  "@context": "http://www.w3.org/ns/odrl.jsonld",
  "@type": "odrl:Set",
  "profile": "https://w3c.github.io/odrl/ai-vocab/",
  "uid": "http://example.org/policy/001",
  "assigner": "urn:example:dataset-owner",
  "target": "http://example.org/asset/dataset1",
  "permission": [{
      "action": "odrl:ai-training",
      "constraint": [{
          "leftOperand": "odrl:purpose",
          "operator": "odrl:eq",
          "rightOperand": "https://w3id.org/dpv#ResearchAndDevelopment"
       }]
    }]
}

The asset (dataset2) maybe used for AI Usage, Search, and Index but must not be used for AI Training.

{
  "@context": "http://www.w3.org/ns/odrl.jsonld",
  "@type": "odrl:Set",
  "profile": "https://w3c.github.io/odrl/ai-vocab/",
  "uid": "http://example.org/policy/002",
  "assigner": "urn:example:dataset-owner",
  "target": "http://example.org/asset/dataset2",
  "permission": [{
        "action": [ "odrl:ai-usage", "odrl:search", "odrl:index" ]
    }],
  "prohibition": [{
      "action": "odrl:ai-training"
    }]
}

The asset (dataset3) maybe used for all Artificial Intelligence actions in Europe, but must not be used for AI Augmentation in Italy.

Note that the conflict property is set to prohibit to ensure that "AI Augmentation" is overridden in the "Artificial Intelligence" action hierarchy.

{
  "@context": "http://www.w3.org/ns/odrl.jsonld",
  "@type": "odrl:Set",
  "profile": "https://w3c.github.io/odrl/ai-vocab/",
  "uid": "http://example.org/policy/003",
  "assigner": "urn:example:dataset-owner",
  "conflict": "prohibit",
  "target": "http://example.org/asset/dataset2",
  "permission": [{
        "action": "odrl:ai",
        "constraint": [{
           "leftOperand": "spatial",
           "operator": "eq",
           "rightOperand":  "https://www.wikidata.org/wiki/Q46"
         }]
    }],
  "prohibition": [{
      "action": "odrl:ai-augment",
      "constraint": [{
           "leftOperand": "spatial",
           "operator": "eq",
           "rightOperand":  "https://www.wikidata.org/wiki/Q38"
        }]
  }]
}

TDM Reservation Example

The TDM Reservation use case centres around the assigner "reserving" their "rights" for "data mining" actions over their asset(s), and to specify that such rights maybe available in consultation with the assigner/owner (in the future).

The owner (assigner) makes an Offer for use of the asset (dataset8) specifically allowing data-mining with a duty (an obligation) to obtain prior consent (from the consentingParty) before the policy can be used (the constraint)."

{
  "@context": "http://www.w3.org/ns/odrl.jsonld",
  "@type": "odrl:Offer",
  "profile": "https://w3c.github.io/odrl/ai-vocab/",
  "uid": "http://example.org/policy/008",
  "assigner": "urn:example:dataset-owner",
  "consentingParty": "urn:example:dataset-licensor",
  "target": "http://example.org/asset/dataset8",
  "permission": [{
      "action": "odrl:data-mining",
      "duty": [{
        "action": "obtainConsent",
        "constraint": [{
                "leftOperand": "event",
                "operator": "lt",
                "rightOperand": { "@id": "odrl:policyUsage" }
            }]
        }]
    }]
}

Introduction

Identifiers