AI Usage Preferences Standardization Standardize interoperable building blocks enabling preferences about the collection and processing of content for artificial intelligence model development, deployment, and use to be expressed and communicated. Submitter's Note: This plan has been inferred by ChatGPT from the charter, work products, meeting materials, and public communications of the Internet Engineering Task Force's AI Preferences Working Group. ^^ It was prompted by Leonard Rosenthol's comment on Renato Iannella's LinkedIn posting concerning the W3C ODRL Community Group's AI Governance Vocabulary: https://www.linkedin.com/posts/riannella_the-w3c-odrl-community-group-has-released-activity-7480448082665500672-s_j-?utm_source=share&utm_medium=member_desktop&rcm=ACoAAACRo-wBsB6AJXaqtz06r_wMwIYVUJtr0PM ^^ The Working Group focuses on attaching preferences to content either through content metadata or through protocols used to deliver content. Its chartered deliverables include a vocabulary for expressing AI-related preferences and protocol specifications for associating those preferences with content. ^^ The AI Preferences vocabulary is intended to enable structured declarations about restrictions or permissions concerning the use of digital assets by automated processing systems. The attachment specification defines mechanisms through which preferences can be signaled when content is acquired using HTTP and may update the Robots Exclusion Protocol to support usage preferences. ^^ The IETF AI Preferences and W3C ODRL AI Governance Vocabulary efforts appear to be complementary. The IETF effort focuses on broadly deployable vocabulary and Internet signaling mechanisms. ODRL can express more detailed policies involving permissions, prohibitions, duties, parties, assets, actions, and constraints. Coordination could reduce semantic conflicts, duplication, and incompatible implementations. ^^ StratML complements both efforts by documenting their intentions, stakeholders, objectives, and prospective results in an open, machine-readable format. AI Preferences Working Group AIPREF f7e53ab7-e448-4a06-96bc-7de7401624f6 Standardizes building blocks enabling preferences about how content is collected and processed for artificial intelligence model development, deployment, and use to be expressed and associated with content. Internet Engineering Task Force Hosts the Working Group and develops voluntary technical standards supporting the design, use, and interoperability of the Internet. AI Preferences Working Group Participants Contribute requirements, proposals, reviews, implementation experience, technical analysis, and consensus to the development of AI preference standards. Mark Nottingham Serves as a co-chair of the AI Preferences Working Group and contributes technical and procedural leadership to its standards-development activities. Suresh Krishnan Serves as a co-chair of the AI Preferences Working Group and contributes technical and procedural leadership to its standards-development activities. Mike Bishop Serves as an Internet Engineering Steering Group area director responsible for oversight of the Working Group. Paul Keller Authors the Working Group's specification defining a vocabulary for expressing AI usage preferences. Martin Thomson Edits and co-authors the Working Group's vocabulary and HTTP attachment specifications. Gary Illyes Co-authors the Working Group's specification for associating AI usage preferences with content in HTTP. Leonard Rosenthol Highlighted the importance of coordinating the Working Group's vocabulary with the W3C ODRL Community Group's AI governance work. Internet Users Depend upon an open and interoperable Internet in which content-related preferences can be communicated and interpreted predictably. An Internet in which preferences concerning the use of content by artificial intelligence and other automated processing systems are expressed and communicated transparently and interoperably. 1ba75b5d-8f9e-4ed1-8c0c-a80b7df76650 To develop Internet standards enabling preferences about the collection and processing of content for artificial intelligence model development, deployment, and use to be expressed, associated with content, and interpreted consistently. ed388e18-5489-4210-b986-d818abb18f70 Interoperability Enable preferences to be expressed, transmitted, discovered, and interpreted consistently across content, protocols, platforms, and automated systems. Openness Develop standards through open participation, public documentation, technical review, and rough consensus. Agency Enable authorized stakeholders to communicate preferences concerning automated uses of content. Transparency Make preference signals and their intended meanings visible and understandable to publishers, implementers, automated systems, and affected users. Simplicity Define building blocks that can be implemented and deployed without unnecessary complexity. Precision Define preference terms and processing behavior clearly enough to reduce ambiguity and inconsistent interpretation. Extensibility Enable preference signaling to evolve as automated processing technologies and legitimate use cases change. Compatibility Integrate preference signaling with established Internet protocols, content formats, metadata systems, and operational practices. Neutrality Provide mechanisms for expressing preferences without determining the legal validity, enforceability, or substantive merits of particular preferences. Vocabulary Standardize a common vocabulary for expressing preferences concerning automated uses of digital assets. c4e90e93-ad11-45af-981f-c5e842661be5 1 Content Creators Use vocabulary terms to express preferences concerning automated uses of their content. Publishers Communicate preferences applicable to content they make available through Internet services. Rights Holders Depend upon sufficiently clear terms to express preferences within the scope of their authority. AI System Developers Implement vocabulary terms in systems that collect, process, train upon, analyze, retrieve, or generate outputs from digital assets. Automated Processing System Operators Interpret preference signals and determine how their systems respond to them. The vocabulary specification defines terms for declaring restrictions or permissions concerning how digital assets are used by automated processing systems. Although motivated by AI, its concepts may apply to other automated processing. The vocabulary should provide protocol-independent semantics that can be used through multiple attachment and signaling mechanisms. Terms Define preference terms representing materially distinct categories of automated content use. d411f4d9-4058-4f83-8c55-21a6b9a521f8 1.1 Vocabulary Editors Draft, refine, document, and maintain the meanings and intended applications of preference terms. Domain Experts Assess whether proposed terms accurately distinguish significant automated processing activities and use cases. Vocabulary terms should represent meaningful distinctions among automated uses rather than particular vendors, products, or rapidly changing implementation techniques. Candidate concepts may address model training, search, retrieval, inference, generation, transformation, and other processing activities, but inclusion should depend upon clear definitions and demonstrated interoperability needs. Semantics Specify protocol-independent meanings for AI usage preference terms. 874d64b6-f180-46ad-962f-9961bf8098d4 1.2 Protocol Designers Apply common preference semantics through different Internet protocols and attachment mechanisms. Implementers Use normative definitions to interpret preference expressions consistently. Separating vocabulary semantics from attachment mechanisms enables the same preference concepts to be used in HTTP signaling, embedded metadata, and other future protocol bindings. Definitions should identify boundaries, overlaps, exclusions, and dependencies among terms. Expressions Define structured forms for declaring permissions, restrictions, or other supported preferences. fbc59100-6ef1-42fd-a918-13bf406f499d 1.3 Preference Authors Create declarations conveying intended restrictions, permissions, or other supported preferences. Preference Consumers Parse and interpret structured declarations when acquiring or processing content. Expression syntax should support predictable machine processing while permitting understandable human presentation. The standard should distinguish absence of a preference from an affirmative permission or restriction and should avoid assigning meanings that cannot be reliably inferred. Examples Provide examples illustrating the intended use and interpretation of preference terms. fd2467e2-655e-48d5-9767-6079faf75c9c 1.4 Specification Readers Use examples to understand the scope, syntax, and intended interpretation of preference declarations. Test Developers Translate representative examples into interoperability and conformance tests. Examples should address simple declarations, combinations of preferences, unspecified uses, conflicting signals, unsupported terms, and realistic content publication and acquisition scenarios. Association Standardize mechanisms for associating AI usage preferences with Internet content. 12a63e2e-f945-45d2-a6ed-2c3c4959360b 2 Web Publishers Attach or signal preferences applicable to resources made available through HTTP. Web Server Developers Implement mechanisms for communicating preference information with HTTP resources. AI Crawlers Discover and interpret preference signals while acquiring Internet content. Content Management System Developers Provide publishers with usable controls for assigning preferences to content. Content Distribution Networks Preserve and communicate applicable preference signals when delivering content. The Working Group's HTTP attachment specification defines how AI usage preferences can be signaled as part of content acquisition. The charter permits preferences to be associated with content through embedded metadata or through the protocol that delivers the content. Attachment mechanisms should preserve the common vocabulary's semantics while addressing protocol-specific discovery, scope, caching, representation, and precedence issues. HTTP Signaling Define mechanisms for communicating AI usage preferences when content is acquired through HTTP. 56d0f552-dddf-43ad-9c5e-1158cc22d8c0 2.1 HTTP Implementers Implement preference signaling in clients, servers, intermediaries, crawlers, and related software. Resource Publishers Associate preference expressions with HTTP resources and representations. HTTP signaling should specify how preferences are located, retrieved, scoped, cached, updated, and associated with resources or representations. It should address the behavior of intermediaries and avoid disrupting established HTTP semantics. Robots Integration Integrate AI usage preference signaling with the Robots Exclusion Protocol where appropriate. 5a398f73-db5a-4999-ba49-91b1ee21de9c 2.2 Web Crawler Operators Interpret applicable preference information communicated through or alongside robots.txt. Website Operators Publish crawler access directives and AI usage preferences through compatible mechanisms. The attachment specification proposes updating RFC 9309 to support inclusion of usage preferences. Access-control directives and downstream usage preferences serve different purposes and should remain distinguishable even when communicated through related infrastructure. Metadata Support association of preferences with content through applicable metadata mechanisms. a9432862-df48-4e92-98b6-2d40585f301d 2.3 Content Format Communities Evaluate mechanisms for embedding or linking preference expressions in content metadata. Metadata System Developers Preserve and expose preference information when content is stored, transformed, distributed, or rendered. Embedded metadata may remain associated with content after acquisition but can be removed, transformed, or separated from the asset. Protocol-level and embedded signals should use compatible semantics and specify how discrepancies are handled. Scope Specify the content, representations, resources, agents, and processing contexts to which preference signals apply. 1940f987-33a9-4ae1-8e52-3b2a711abdfa 2.4 Publishers Determine the intended scope of preferences they associate with content. Preference Processing System Developers Apply scope rules when determining which preference expressions relate to acquired content. Scope rules should address individual resources, collections, sites, representations, embedded components, linked assets, temporal applicability, and other circumstances in which a signal might otherwise be interpreted too broadly or narrowly. Interoperability Promote consistent interpretation and implementation of AI usage preferences across Internet protocols, platforms, content formats, and automated processing systems. 5f52ef0e-96c2-46e2-a4cc-18f9d3ebad1f 3 Standards Organizations Coordinate related specifications, vocabularies, protocols, metadata, and implementation guidance. Platform Providers Implement interoperable mechanisms for publishing, discovering, communicating, interpreting, and applying AI usage preferences. Browser Developers Evaluate opportunities for exposing AI usage preference information to users and Web applications. Search Engine Providers Determine how AI usage preferences should influence crawling, indexing, retrieval, and other automated processing activities. Regulators Consider whether standardized AI usage preferences can facilitate transparency and compliance without creating new legal obligations. Interoperability requires agreement on semantics as well as syntax. Different implementations should interpret equivalent preference expressions consistently while documenting any limitations, extensions, or unsupported features. Consistency Promote consistent interpretation of AI usage preference expressions across implementations. 7df4560e-4c5d-4b95-b4f9-4aebd2f7d76c 3.1 Software Implementers Develop interoperable processing behavior for supported preference expressions. Conformance Test Developers Create tests demonstrating consistent implementation of the specifications. Implementations should avoid assigning incompatible meanings to identical preference expressions and should disclose any implementation-specific extensions. Coordination Coordinate AI usage preference standards with related standards organizations and technical communities. 1c1fd6b5-f937-4f17-bb73-649c41bb357f 3.2 World Wide Web Consortium Develops complementary standards including the ODRL AI Governance Vocabulary. International Press Telecommunications Council Develops metadata standards widely used for digital publishing. PLUS Coalition Develops standards supporting licensing and rights management for media assets. Leonard Rosenthol Encouraged coordination between the AI Preferences Working Group and the W3C ODRL Community Group. The AI Preferences Working Group charter explicitly calls for liaison with other standards organizations where appropriate. Coordination should minimize unnecessary duplication while promoting compatible semantics and complementary capabilities. Mappings Document semantic relationships between AI usage preferences and related vocabularies. 2a5dc83d-fac8-4608-81d8-d523e90fca32 3.3 Vocabulary Maintainers Identify equivalent, broader, narrower, related, and incompatible concepts among standards. Knowledge Graph Developers Use documented mappings to improve semantic interoperability among standards and implementations. Mappings should distinguish semantic equivalence from mere lexical similarity and should identify cases where concepts only partially overlap or differ materially. Reconciliation Specify methods for reconciling multiple applicable AI usage preference expressions. 8f2dfe59-f8c9-4b80-a6fb-b72526d6198b 3.4 Preference Processing System Developers Implement consistent methods for resolving multiple applicable preference expressions. Publishers Benefit from predictable handling when preferences are expressed through multiple mechanisms. The charter specifically includes reconciliation of preferences as part of the Working Group's scope. Reconciliation rules should clearly document precedence, conflict handling, inheritance, and situations in which preferences cannot be reconciled automatically. Implementation Support practical implementation and deployment of interoperable AI usage preference standards. 28d6c113-50f7-4669-a2af-59fe9234d0b8 4 Software Developers Develop software supporting publication, discovery, interpretation, and application of AI usage preferences. AI Service Providers Determine how their systems discover, interpret, document, and respond to applicable AI usage preferences. Content Publishers Deploy preference signaling mechanisms on production Web sites and services. Open Source Communities Develop reusable implementations supporting widespread adoption of the standards. Successful implementation depends upon more than publication of technical specifications. Practical deployment requires documentation, reference implementations, testing, interoperability demonstrations, operational experience, and continuing maintenance. Guidance Publish implementation guidance supporting consistent deployment of the specifications. 88d1fb59-b0b1-4724-bc76-67e2b5b6d65e 4.1 Implementation Guide Authors Develop practical guidance supporting interoperable implementation of AI usage preference standards. Implementers Apply guidance when deploying AI usage preference capabilities in production systems. Implementation guidance should explain vocabulary usage, attachment mechanisms, processing expectations, error handling, security considerations, caching behavior, extensibility, and interactions with existing Internet protocols. Guidance should distinguish normative requirements from implementation recommendations. Reference Implementations Encourage development of reference implementations demonstrating interoperable use of the specifications. cc2dce2f-5dbd-4f69-b9d6-87eb15bda4a8 4.2 Open Source Developers Develop reusable software illustrating implementation of AI usage preference standards. Commercial Software Developers Incorporate interoperable AI usage preference capabilities into commercial products and services. Reference implementations help clarify ambiguous requirements, validate specification design decisions, accelerate adoption, and provide practical examples for future implementers. Testing Develop interoperability tests demonstrating consistent implementation of the standards. a4898f73-5144-44c4-bfd6-c67d45b83dc3 4.3 Test Developers Create interoperability and conformance tests covering representative implementation scenarios. Independent Implementers Demonstrate consistent interpretation of AI usage preferences through independently developed implementations. Testing should include valid and invalid preference expressions, conflicting declarations, unsupported vocabulary terms, multiple attachment mechanisms, caching behavior, precedence rules, and realistic operational scenarios. Interoperability claims should be supported by objective evidence. Deployment Promote deployment of interoperable AI usage preference capabilities across the Internet. 2c6d615d-1f2d-4585-bfb0-596e05d39d36 4.4 Website Operators Deploy AI usage preference signaling for publicly accessible Internet resources. Cloud Service Providers Support AI usage preference mechanisms within hosted Web infrastructure and content delivery platforms. AI Platform Operators Document how AI usage preferences influence acquisition and processing of Internet content. Deployment should encourage transparent documentation of supported preference expressions, implementation limitations, operational behavior, and changes over time so that publishers and users understand how preference signals are actually processed. Governance Maintain and evolve AI usage preference standards through open Internet standards processes. 00cfde5e-1d4d-4d92-a847-b3b3b4ef7bf5 5 Working Group Participants Review proposals, resolve issues, evaluate implementation experience, and build technical consensus. Public Reviewers Identify ambiguities, omissions, implementation concerns, and opportunities for improvement. Internet Engineering Steering Group Reviews Working Group deliverables and determines their readiness for publication as RFCs. Internet Architecture Board Provides architectural oversight and coordination across Internet standards activities. The IETF standards process emphasizes open participation, technical merit, running code, interoperability, and rough consensus. AI usage preference standards should evolve through documented experience, technical evaluation, and broad community review rather than through unilateral decisions. Participation Encourage broad participation in development and review of AI usage preference standards. 31df2a35-c44f-4b84-b77e-c6575d170fa8 5.1 Content Publishers Contribute operational experience concerning publication and communication of AI usage preferences. AI Service Providers Contribute implementation experience concerning interpretation and application of AI usage preferences. Researchers Evaluate emerging requirements, implementation experience, and future directions for AI usage preference standards. Broad participation improves technical quality by incorporating operational experience from publishers, AI developers, Internet infrastructure providers, standards organizations, researchers, and other affected communities. Review Evaluate proposed specifications through open technical review and documented issue resolution. 48fd4c79-2e95-4c90-a689-662e5c9e95a4 5.2 Specification Reviewers Assess clarity, completeness, consistency, interoperability, implementability, security, privacy, and extensibility. Document Editors Document issues, proposed resolutions, revisions, and remaining questions throughout standards development. Technical review should distinguish implementation questions from legal, policy, contractual, or ethical questions that may depend upon jurisdictions or organizational policies rather than Internet standards. Evolution Revise AI usage preference standards in response to implementation experience and evolving Internet technologies. 55bce663-27dc-4fc5-b28f-c53f23675c55 5.3 Specification Editors Manage revisions, versioning, corrections, extensions, and documentation of the standards. Implementers Provide operational experience identifying ambiguities, implementation challenges, and opportunities for improvement. Future revisions should preserve stable semantics whenever practical, document compatibility impacts, and provide migration guidance when changes affect deployed implementations. Accountability Document decisions, implementation experience, and evidence supporting development and deployment of the standards. e6a99d0f-dfd0-4af9-a40e-07f9fb5e726d 5.4 Working Group Chairs Guide the standards-development process, document consensus, and coordinate completion of Working Group deliverables. Standards Implementers Document supported capabilities, implementation limitations, and interoperability experience. Future rendition of this plan in StratML Part 2 format could associate performance indicators with objectives such as Working Group participation, issue resolution, interoperable implementations, deployment of preference signaling, publication of RFCs, liaison activities with other standards organizations, and adoption by publishers and AI service providers. 2025-12-09 2026-07-10 https://datatracker.ietf.org/wg/aipref/about/ Owen Ambur Owen.Ambur@verizon.net