ISRC-AAM-CID 2.0: Towards an Interoperable Global Trust Infrastructure for AI Music

Published by the Australian AI Music Alliance and OZBeat AI, this Discussion Draft builds on and advances the ISRC-AAM-CID 1.0 framework.

Abstract

In February 2025, the Australian AI Music Alliance published “ISRC-AAM-CID: Establishing a Global Standardized AI Music Identifier”, introducing ISRC-AAM-CID 1.0 as a framework for identifying and disclosing the use of AI in music. At the time, the music industry already had well-established identifier systems for musical works, sound recordings, performers and rights-related parties. Generative AI introduced a new challenge: how to record whether AI was used in creating a piece of music, where and how it was used, the extent of its involvement, and the specific digital content to which that information relates.

Version 1.0 addressed this challenge through a three-layer structure. ISRC identified the Sound Recording; AAM (AI Application Music) described the degree of AI involvement across music creation and production; and CID provided a content-addressed link between the registration record and the specific digital content submitted.

Since then, the framework has moved from concept into practical implementation. The Alliance has established an AI music registration and lookup system and applied the framework to more than 1,600 registrations, over 200 AI music showcases and competitions, eight AI music charts, and a growing range of commercial and industry use cases. This experience has reinforced the practical need for greater transparency around AI use in music, while also revealing where the 1.0 framework can evolve to support broader interoperability across the industry.

The wider music ecosystem has also moved rapidly. In September 2025, Spotify announced support for an AI disclosure standard being advanced through DDEX, describing AI use as a “spectrum, not a binary”. In July 2026, IFPI, RIAA, A2IM, WIN, IMPALA, The Recording Academy, SAG-AFTRA and the Human Artistry Campaign jointly introduced voluntary “AI-Generated” and “AI-Assisted” labels for sound recordings. C2PA Content Credentials and the Creator Assertions Working Group (CAWG) have continued to advance mechanisms for provenance, identity, training and data-mining assertions. Transparency obligations under Article 50 of the EU AI Act began to apply on 2 August 2026. On 31 July 2026, the Munich Regional Court I issued a significant ruling in GEMA v. Suno concerning unauthorised reproductions for training, reproductions within the model, and related outputs.

Together, these developments point to a new phase for the industry. The question is no longer simply whether AI-generated or AI-assisted music should be labelled. The more important questions are becoming: How granular should AI disclosure be? How can it be made machine-readable? How should Musical Work identity, Sound Recording identity, AI disclosure, digital files, provenance and identity verification be connected? And how should the respective roles of different standards and infrastructure be defined so that they can interoperate effectively?

Drawing on more than a year of practical implementation and the rapid evolution of the industry, this Discussion Draft proposes six areas of evolution for ISRC-AAM-CID 2.0: moving from a serial composite structure to a parallel architecture better suited to interoperability; evolving AAM weighting from a static model towards dynamic, versioned contribution profiles; transforming free-text disclosure into structured and, where meaningful, quantifiable data; using CID as a content anchor for a broader File Identity Record; developing a Party and Representation model that can distinguish natural persons, digital representations and virtual identities while separating Party from Contribution; and extending AAM beyond the ISRC-identified Sound Recording layer to the ISWC-identified Musical Work layer.

Together, these developments move ISRC-AAM-CID towards a broader AI Music Identity & Trust Architecture: an open, extensible and interoperable framework built on established music identifiers, with AAM providing a music-specific AI disclosure layer and the File Identity Record connecting music entities to specific digital assets. The 2.0 direction also explores interoperability with the DDEX music data exchange ecosystem and C2PA provenance and authenticity infrastructure, with the aim of enabling AI disclosure, music identity and digital content provenance to move together more clearly and reliably across the music value chain.

1. Why We Proposed ISRC-AAM-CID in 2025

The music industry has built a rich ecosystem of identifiers. ISWC identifies a Musical Work, ISRC identifies a Sound Recording, and systems such as IPI, IPN and ISNI identify different types of creators, performers, rights-related parties and public identities. As discussed in 1.0, these identifiers and their associated metadata underpin the digital music business at scale: works can be identified, recordings tracked, contributors and their contributions accurately recorded, usage data reported, and revenues routed through complex value chains to the relevant parties.

Generative AI added a further question: when AI is used in one or more stages of lyrics, melody, vocals, instrumentation, arrangement, production, mixing or mastering, how should that involvement be described accurately?

AI involvement can take many forms. A creator may use AI only for brainstorming, or generate an accompaniment with AI before having it replayed by human musicians. Human-written lyrics may be paired with synthetic vocals. At the other end of the spectrum, a prompt may generate a near-complete recording that is then selected, edited and post-produced by people. These approaches reflect different creative processes and may carry different implications for consumer expectations, platform policies and potential rights issues.

Version 1.0 therefore sought to make AI involvement recordable. AAM0-AAM4 condensed complex information about AI participation into five readily understandable levels, while CID created a stable link between that disclosure and the specific digital content submitted at registration.

2. From Proposal to Operation: What 1.0 Actually Taught Us

The value and opportunities of a standard become much clearer once it is used in practice. Over the past year, ISRC-AAM-CID has moved beyond the white paper into operational workflows for registration, lookup, presentation and commercial and industry applications. At the time of this Discussion Draft, the Alliance has recorded more than 1,600 AI music registrations, hosted over 200 AI music showcases and competitions, and operated eight AI music charts. ISRC-AAM-CID has also been applied across music presentation and discovery, marketplace transactions and broader AI music community initiatives.

Practical use has brought increasingly concrete requirements into view. Creators want to show that their work was not simply generated at the press of a button and to describe human-AI collaboration more accurately. Labels are interested in batch registration and compatibility with existing supply chains. Buyers want traceable information about the works they acquire. And when files are transcoded, re-exported or delivered to different platforms, the relationship between the registration record and each specific file needs to remain clear.

These experiences have sharpened our view of the long-term value of ISRC-AAM-CID. It began as a framework for AI music identification and disclosure, but it has also become a practical lens for understanding what AI-related information the music industry needs to record, how that information should be structured, and how it can move among creators, labels, platforms, buyers and rights-related parties.

2.1 1.0 Showed the Need for an Intermediate Layer of AI Disclosure

In practice, we have seen the limits of two common approaches to disclosure. Free text can preserve rich detail about the creative process, but it is difficult to standardise and exchange. A simple “AI / non-AI” label is easy to display and understand, but it removes much of the information about how AI was actually used.

What is missing is an intermediate layer that preserves meaningful detail while remaining understandable and exchangeable. AAM is intended to provide that layer. It brings together creative roles, AI-use activities and degrees of involvement into a concise composite result, while preserving the ability to inspect the underlying detail when needed.

2.2 1.0 Clarified the Boundaries Between Different Information Layers

For ease of understanding and communication, 1.0 presented ISRC, AAM and CID in a serial ISRC-AAM-CID form. This made the relationship among recording identity, AI involvement and digital content immediately visible and proved useful as a practical expression within the Alliance.

As adoption expanded, the distinct roles of these three information types became clearer. ISRC identifies the recording, AAM describes AI involvement, and CID anchors the record to specific digital content. They are related, but each has its own data structure and lifecycle.

Version 2.0 therefore proposes a more explicit separation of these layers. It preserves the intuitive presentation of 1.0 while allowing each layer to be managed independently and connected through defined references, providing a clearer foundation for interoperability.

3. Over the Past Year, the Industry Has Begun to Converge on Transparency, Exchangeability and Verifiability

At the beginning of 2025, the industry was still widely debating whether AI music required special identification. By 2026, the discussion had advanced. Industry bodies, platforms, standards organisations and regulators were focusing on more concrete questions: how information about AI involvement can be understood by consumers, used by platforms, exchanged through the supply chain, detected by technical systems and, where necessary, support provenance and rights tracing.

Together, these developments form the industry context for 2.0. AI transparency is moving from policy discussion into the infrastructure of the music ecosystem.

3.1 DDEX and Spotify: AI Use Is Not a Binary

In September 2025, Spotify announced support for an industry AI disclosure standard being advanced through DDEX and stated that artists and rights holders should be able to indicate where AI played a role in a recording. Spotify described AI use as a spectrum rather than a binary, closely reflecting what we had observed through the operation of 1.0. Public DDEX materials likewise indicate ongoing work on how generative-AI involvement can be communicated more effectively across the music value chain.

This points to an increasingly clear industry direction: meaningful AI transparency needs to communicate both where AI was involved and the nature of that involvement.

3.2 IFPI / RIAA and Others: From Back-End Metadata to Consumer-Facing Labels

On 10 July 2026, IFPI, RIAA, A2IM, WIN, IMPALA, The Grammys, SAG-AFTRA & Human Artistry Campaign jointly introduced a voluntary track-level labelling approach for sound recordings, using “AI-Generated” and “AI-Assisted” to distinguish different degrees of generative-AI use. The approach intentionally uses simple, high-level labels and currently applies to sound recordings, while explicitly excluding lyrics, compositions, music videos and cover art.

This development also helps clarify the role of AAM. Consumer labels and industry-level disclosure can operate at different levels of granularity. A consumer may only need to know that a recording is “AI-Assisted”; a label, platform, rights holder or researcher may need to know whether AI was involved in vocals, composition, instrumentation or production. AAM can provide that more granular layer while complementing simpler consumer-facing labels.

3.3 C2PA and CAWG: From “What Was Declared” to “Who Declared It, and What Asset It Relates To”

C2PA Content Credentials continue to evolve around digital content provenance and verifiable assertions. C2PA 2.3 expands capabilities for linking assets and external information. The Creator Assertions Working Group (CAWG) is also advancing specifications relating to identity, metadata, training and data mining, enabling a digital asset to communicate not only what happened, but also who made a declaration and what asset that declaration concerns.

These developments have prompted us to reconsider the role of CID in 2.0. Content addressing can verify whether digital content remains consistent, while a complete file identity record requires additional context, including identity, time, declarations and provenance. Version 2.0 therefore draws a clearer distinction between a content fingerprint and a File Identity Record, while exploring alignment with mature provenance infrastructure.

3.4 The EU AI Act: Transparency Is Entering Regulatory Infrastructure

The relevant transparency obligations under Article 50 of the EU AI Act began to apply on 2 August 2026. Related European Commission guidance states that certain AI systems must provide machine-readable marking for AI-generated or manipulated content in support of detection and transparency requirements. Machine-readable AI marking is therefore beginning to enter the compliance architecture of some regulatory environments.

This creates a clear long-term design requirement for ISRC-AAM-CID: AI music disclosure should support machine readability, interoperability, robustness and verifiability from the outset. AAM can contribute a disclosure layer designed specifically for the music industry.

3.5 GEMA v. Suno: Provenance, Authorisation and Training Sources Enter More Concrete Evidentiary Contexts

On 31 July 2026, the Munich Regional Court I issued its judgment in GEMA v. Suno (42 O 763/25), concerning six musical works. The decision addressed issues including unauthorised reproduction for training purposes, storage of works within the model and related outputs; GEMA described the ruling as a significant victory for music creators. The case brings questions of training sources, licensing, reproduction within models, and the relationship between outputs and existing works into a more concrete setting of evidence and legal remedies.

For 2.0, this further highlights the value of traceable records. A disclosure record with long-term utility should be able to show who made the declaration, what was declared, the basis for it, which Work or Recording it relates to, which digital asset it concerns, and which aspects of the information have been verified.

3.6 Our Assessment: The Industry Is Moving from “Whether to Disclose” to “How to Establish Trust”

Taken together, these developments reveal three increasingly clear and interconnected tracks: metadata exchange across the music supply chain, consumer-facing labelling, and provenance and authenticity for digital assets.

Each addresses a different question: how information moves, how it is understood, and how it can be traced and verified. As these tracks mature, the central challenge for AI music is changing as well. The industry is beginning to establish disclosure mechanisms; the next question is how those disclosures can be connected and become usable, trusted records across different participants and systems.

This is the context in which ISRC-AAM-CID is moving into 2.0. Version 1.0 established the basic relationship among recording identity, AI involvement and digital content. Version 2.0 builds on that foundation by adding greater structure, traceability and interoperability, and by exploring alignment with music data exchange, consumer-labelling and digital content provenance systems.

4. Reframing AAM + CID: Addressing the Trust Problem in the AI Era

Experience with 1.0 has made the long-term value of AAM + CID increasingly clear. At its core is a fundamental issue: information asymmetry and the trust cost it creates. Creators usually know the AI-assisted creative process best. Platforms and labels often receive the work with limited information. Consumers encounter only the finished music. When disputes arise, rights holders may need to trace the relevant Work, Recording, file, timing and declarations.

AAM + CID is intended to reduce that information gap. AAM turns AI involvement into structured disclosure; CID connects that disclosure to specific digital content. Together, they allow information that was previously embedded in the creative process to be recorded, carried forward and, where appropriate, traced and verified.

4.1 AAM + CID Responds to Trust Asymmetry

AAM is designed to remain neutral in this context. Its role is to record how AI participated in creation and production and to turn information that was previously dispersed or implicit into portable, interpretable data. Different participants can then apply their own policies, contracts and legal frameworks.

4.2 AAM-Verified: From Structured Disclosure to a Verifiable Trust Record

As disclosure becomes more structured, the next question is the degree of confidence that can be placed in it. Three layers can be distinguished more clearly:

AAM is a disclosure language for AI involvement.

AAM-CID links that disclosure to specific digital content, creating a traceable structured record.

AAM-Verified adds defined-scope verification, providing users with a higher level of assurance about specified aspects of the record.

An open standard allows different participants to speak the same language. Verification can then indicate which aspects of a record’s source, supporting evidence, identity information and digital-asset binding have been checked.

The meaning of AAM-Verified should always correspond to the actual scope of verification. A record might confirm that the source of a declaration is traceable, that supporting evidence has been recorded, that a digital asset has been cryptographically bound, or that particular identity information has been verified. As 2.0 develops, these verification dimensions, methods and levels can be standardised further.

4.3 AAM as an Extensible Layer: Beyond ISRC to Other Music Entities

Version 1.0 began with ISRC because the Sound Recording is the most direct object through which AI music enters distribution and consumer markets. With deeper operational experience, we have seen that the AI involvement described by AAM can exist across different music entities and contributions, giving it a basis for broader extension.

AAM can operate alongside ISRC to describe AI involvement at the Sound Recording level, and alongside ISWC to describe AI involvement at the Musical Work level. In more granular applications, AAM may also be used to describe AI involvement within a specific Contribution.

This allows AAM to evolve into a reusable AI disclosure layer. It can operate alongside established identifiers and continue to adapt as the industry’s requirements for AI disclosure granularity develop.

4.4 Working with DDEX and C2PA: Letting Each Standard Play to Its Strengths

ISRC-AAM-CID 2.0 will prioritise collaboration with mature industry infrastructure. DDEX has long supported standardised data exchange across the digital music value chain, while C2PA provides an open technical foundation for digital content provenance and authenticity. Version 2.0 will further clarify the role of AAM in AI disclosure and the role of CID and File Identity, while exploring effective alignment with these ecosystems.

Such collaboration allows different standards to retain clear responsibilities: music identifiers identify works and recordings; AAM describes AI involvement; File Identity connects specific digital assets; provenance infrastructure provides source and verification context; and industry data-exchange systems help that information move through the supply chain.

5. Six Directions for the Evolution of 2.0

Based on the practical experience and observations above, we propose six interconnected directions for evolution. Together they form the principal path from ISRC-AAM-CID 1.0 to 2.0: refine the overall architecture; make AAM more representative of real creative processes; progressively structure disclosure; use CID as the content-addressing foundation for a File Identity Record; and extend the framework to Party and Musical Work layers.

These six directions arise from the operation of 1.0 and respond to the music industry’s growing need for transparency, traceability and interoperability around AI-related information.

5.1 Evolution One: From a Serial Composite to a Parallel Architecture for Clearer Interoperability

Version 1.0 used a serial ISRC-AAM-CID form, presenting the recording identifier, AAM level and CID together in a human-readable composite code. The format was intuitive, easy to communicate, and helped users quickly understand the relationship among the three information layers.

As adoption expands, Version 2.0 will define the relationship among these three information layers more explicitly at the data-structure level. ISRC retains its established structure and industry semantics, while AAM-CID exists independently as an associated enriched information layer. For user-facing purposes, the three layers can still be presented together; at the data level, however, they are recorded separately and linked through explicit references.

This evolution preserves the accessibility and adoption advantages of 1.0 while giving 2.0 a clearer data architecture and greater room for interoperability.

The parallel architecture also better reflects the real-world lifecycle of music data. A Sound Recording can retain the same ISRC over time, while its associated AAM disclosure may evolve as additional information is provided or new versions of the disclosure are issued. The same recording may also be associated with multiple digital files, such as a WAV master, a platform delivery file, a transcoded MP3 or a preview, each with its own content fingerprint and version relationships.

By managing these layers separately, recording identity, AI disclosure and specific digital files can retain their own stability while remaining connected through explicit relationships. This provides a clearer structural foundation for future alignment with music data exchange, digital content provenance and other industry infrastructure.

5.2 Evolution Two: From Static to Dynamic AAM Weighting — From a Fixed Model to a Versioned Creative Profile

Version 1.0 assigned default weights to typical roles including composer, lyricist, producer, vocalist, instrumentalist, recording engineer, mixing engineer and mastering engineer, and combined those weights with the degree of AI involvement in each role to calculate an overall AAM result. The model created a common computational baseline and gave a quantifiable, comparable expression to what had previously been a relatively subjective idea of “how much AI was used”.

As the framework has been applied more widely, we have seen significant variation in the creative structures of different music projects. Instrumental music, rap, electronic music, screen music, virtual-artist projects and live recordings can place very different emphasis on particular creative roles. In commercial projects, collaboration agreements, production records and contracts may also provide further evidence of how contributions are actually distributed among the parties involved.

For this reason, 2.0 proposes a dual-track model. A public, versioned Default AAM Profile would remain as a common baseline across projects. Where there are clear creative records, collaboration agreements or other reasonable grounds, a Contextual Contribution Profile could be used to represent the actual structure of a particular work or recording more accurately.

The value of dynamic weighting lies in the additional context it provides. When default weights are adjusted, the associated record should also identify who made the declaration, how the weights were allocated, the basis for the adjustment, and which version of the calculation rules was used. This preserves the flexibility of the model while ensuring that the resulting AAM output has clear provenance and remains traceable.

The design also makes AAM more suitable for different scales of use. Individual creators can use the Default AAM Profile to register with minimal friction. Labels, distributors and production companies can combine more complete session data, production records, contract data and batch catalogue workflows with a Contextual Contribution Profile, while remaining within the same AAM semantic framework.

5.3 Evolution Three: From Free-Text Disclosure to Structured and Quantifiable Disclosure

The 1.0 registration process retained substantial free-text input, including which AI tools were used, the relevant usage-rights tier, where AI participated, and how the creative process unfolded. At an early stage, when terminology and practices were still changing rapidly, this approach captured a wide range of real-world use cases.

As registration volumes grew, the limitations of free text became clearer. The same tool may appear under different names or spellings, and the same creative activity may be described in entirely different language by different applicants. Such information may be readable to people, but it is difficult to normalise, compare, aggregate, verify or process by machines.

Version 2.0 therefore proposes a gradual shift from Narrative Disclosure to Structured Disclosure. Information that can be described objectively should, where possible, be captured through structured fields. Elements with clear quantitative meaning can be measured more precisely, while free text should remain available to explain complex human-AI workflows and exceptional circumstances. This approach preserves the creative context of a disclosure while making the underlying data more consistent and easier to exchange.

This evolution can be summarised as:

Narrative → Structured → Machine-readable → Interoperable

The AAM level sits above this information structure as a concise summary of complex AI involvement. More detailed structured data remains available underneath for further review, exchange and verification. In this way, an AAM record can provide a simple, accessible expression for consumers and general users, while also making more granular information available to labels, platforms and other industry participants.

Structured disclosure also creates a stronger foundation for use at scale. Platforms can apply their own presentation or governance rules to the data; labels can analyse AI usage across a catalogue; and creators can reuse the same structured information across different downstream services, reducing the need to repeatedly re-enter or re-explain their creative process.

5.4 Evolution Four: From CID to a File Identity Record: Making the Content Fingerprint the Foundation of File Identity

In 1.0, CID used content addressing to create a stable link between a registration record and specific digital content. When the content of a file changes, its CID changes accordingly, allowing the system to test whether the current content matches what was registered. This remains an important foundation of 2.0.

As operational use deepened, we saw more clearly that “identity” in music exists at several levels. The identity of a recording, the content fingerprint of a file, the identity of a specific file, and the provenance and history of that file answer different questions.

CID provides an important foundation for File Identity. It binds a record to exact digital content and can reveal whether that content has changed. Adding the file’s relationship to a recording, its role, registration time, declaring party, version relationships and provenance references can create a more complete File Identity Record.

Version 2.0 therefore proposes using CID / hash as the content anchor, with music-related identity and relationship information structured around that anchor.

This gives a digital file richer context than a content fingerprint alone. CID answers “what is the content?”; a File Identity Record goes further and asks “which file is this, and how does it relate to a music entity, a declaring party and a provenance record?”

This raises a question worth broader industry discussion. The music industry already has mature Work Identity and Recording Identity systems, yet music is produced, delivered, distributed and archived as concrete digital files. A single recording may correspond to a master, delivery files, previews, transcodes in multiple formats and later derivatives.

We therefore propose exploring whether the industry needs an exchangeable File Identity layer that can reliably connect specific digital files to established Work Identity, Recording Identity and provenance infrastructure. If such a layer is useful, what information should it contain, and how should it work with existing standards?

Version 2.0 treats the File Identity Record as a key area for exploration and invites industry discussion and practical testing to determine its value and appropriate structure.

5.5 Evolution Five: From “Human or AI” to a Clearer Party and Representation Model

AI music is introducing more varied forms of participation and identity. A natural person may use AI to assist creation, or authorise the use of their voice or likeness in a digital representation. A virtual artist may operate under an independent name and visual identity, maintain an ongoing catalogue, and develop a recognisable public identity.

These developments make “who participated?” and “how was the content created?” two distinct questions. Version 2.0 therefore considers distinguishing at least Natural Person, Digital Representation / Digital Twin, and Synthetic / Virtual Identity, while treating Party and Contribution as related but independently recorded dimensions.

This distinction allows AI music participation to be described more accurately. A contribution may be generated or heavily assisted by AI while the credited Party remains a real creator, performer or producer. Conversely, a virtual artist’s output may contain substantial human creative and production contributions.

Party therefore answers the question of “who participated, and in what identity or capacity?”, while Contribution answers “who was responsible for the specific contribution, and how was it created?”. Recording these two dimensions separately provides a clearer way to represent the evolving relationships among natural persons, digital representations, virtual identities and AI-assisted creation, while leaving room for future alignment with established music identity and contributor systems.

5.6 Evolution Six: From ISRC to ISWC: Extending from the Recording Layer to the Work Layer

Version 1.0 used ISRC as its primary entry point because the Sound Recording is the most direct object through which music enters distribution, playback and market transactions. As AI music practices have developed, however, it has become clear that many important uses of generative AI occur before the recording is created. Activities such as lyrics and composition first belong to the Musical Work layer.

Distinguishing the Musical Work from the Sound Recording is therefore important for accurate AI disclosure. The same Musical Work can give rise to multiple Sound Recordings, each using AI differently in vocals, instrumentation, production, mixing and other recording-stage activities, while sharing the same work-level history of AI involvement.

Similarly, a Musical Work created primarily by humans may subsequently be produced as a Sound Recording with extensive AI involvement. Work-level AI involvement and Recording-level AI involvement should therefore be recorded separately, with clear relationships established between them.

This extension begins to form a cross-layer AI Music Identity Architecture. ISWC + AAM can describe AI involvement at the Musical Work level; ISRC + AAM can describe AI involvement at the Sound Recording level; and the File Identity Record can connect those music entities to the digital files that actually circulate.

Within this structure, AAM can serve as a reusable AI disclosure layer that works alongside different types of music identifiers. Musical Works, Sound Recordings and digital files retain their own distinct identity systems, while remaining connected through AAM disclosure and defined relationships. Together, these layers provide a multi-level information structure for AI music, spanning the creative process, the recording, and the specific digital assets associated with them.

6. From Six Evolution Directions to a More Complete AI Music Identity Architecture

Taken together, the six directions for evolution begin to reveal the broader shape of ISRC-AAM-CID 2.0: a layered AI Music Identity Architecture.

This architecture establishes distinct information layers for music identity, AI disclosure, participant identity, file identity, provenance and verification, and industry data exchange. Each layer addresses a different question and, wherever possible, is designed to work with established music-industry standards and digital-content infrastructure.

The core principle is clear separation with explicit relationships. ISWC and ISRC identify works and recordings; AAM describes AI involvement; Party / Representation expresses participants and identity relationships; CID / hash anchors specific digital content; the File Identity Record adds file-level identity and relationship information; provenance infrastructure such as C2PA provides source and verifiable-assertion context; and industry data-exchange systems help relevant information move through the music supply chain.

This layered approach allows each type of information to retain clear semantics and its own lifecycle while forming a complete record through defined links. Existing standards continue to perform the functions for which they are already mature. AAM-CID focuses on the emerging disclosure and file-identity requirements created by AI music and is intended to complement those systems.

Under this architecture, AAM-CID can be referenced, exchanged, verified or displayed according to the needs of a particular use case. Consumers may see concise AI-involvement information, while creators, labels, platforms and other industry participants may access more detailed structured disclosure, file relationships and provenance information.

7. AAM-Verified: From an Open Standard to a Verifiable Trust Layer

As 2.0 develops an open AAM vocabulary, weighting profiles, Structured Disclosure and File Identity Record, records created under the same standard may still rest on different levels of evidence. Some information may be self-declared by creators or rights-related parties; other information may be checked against identity, files, evidence or provenance.

AAM-Verified is intended to express that distinction. Any conforming implementer may generate an AAM Record under the open standard. AAM-Verified indicates that a verification entity has applied published rules to verify defined aspects of the record within a stated scope.

Verification can be performed across different dimensions, including declarant identity, cryptographic binding of the file, supporting evidence, provenance information, and supporting documentation relating to authorisation. Each verification should state what was checked, by what method and within what scope, so that users can understand both what has been verified and what that verification supports.

As use cases develop, AAM-Verified may support different verification scopes or assurance levels to reflect the needs of creator registration, label management, platform integration, marketplace transactions and other B2B applications.

This structure also creates a clear boundary for the long-term operation of an open standard. AAM vocabulary, data structures and core rules can remain open, while services around the standard, including registration, verification, repositories, enterprise integration and compliance support, can develop sustainably in response to market needs.

The open standard answers “how do we describe this in a common language?” AAM-Verified goes further and answers “which information in this record has been verified?”

8. Working with the Global Standards Ecosystem: Making Information Flow

The long-term value of ISRC-AAM-CID 2.0 depends on whether this information can enter the real music value chain and work with established identifiers, data-exchange systems and digital content provenance infrastructure. Interoperability is therefore a core design direction for 2.0, with particular attention to alignment with established ecosystems such as DDEX and C2PA.

8.1 Working with DDEX: Bringing AI Disclosure into the Music Supply Chain

DDEX has long provided standardised data exchange infrastructure for the global digital music industry, allowing information about works, recordings, participants, rights and commercial activity to move among different participants in the music supply chain.

As AI disclosure evolves from high-level labels towards more granular structured information, a new question emerges: which AI information needs to travel with music data, and how should it connect to established music entities and data relationships?

For AAM, several questions merit further discussion. Which AI disclosures should travel directly through the music supply chain, and which are better accessed through external references? How should AI information at Work, Recording, Contribution and Party levels retain clear semantics and relationships? As File Identity and provenance become more important in digital-content infrastructure, should references to those layers also form part of cross-system connections?

These questions remain open in this Discussion Draft. Our objective is to explore how AAM can preserve its own disclosure semantics while connecting effectively with mature music data exchange systems, allowing information about AI involvement to be read, transmitted and used by different participants.

8.2 Working with C2PA: Creating Verifiable Links Between Disclosure, Digital Assets and Declarants

C2PA provides an open, cross-media technical foundation for digital content provenance and authenticity. For ISRC-AAM-CID 2.0, an important area of exploration is how AAM disclosure, the File Identity Record and Content Credentials can complement one another without duplicating their respective functions.

Within such a division of responsibilities, hash / CID can anchor specific digital content; the File Identity Record can establish relationships between the file, the relevant work or recording and related declarations; AAM can describe AI involvement in music creation and production; and C2PA Content Credentials can provide provenance, assertions, signatures and related verification context.

When these capabilities are connected, an AI music record can provide richer context: who declared the AI involvement; which Musical Work or Sound Recording the declaration concerns; which specific digital file it relates to; whether the file still matches the registered content; whether verifiable provenance exists; and which aspects of the information have been checked.

This collaboration also helps clarify the role of AAM-CID. AAM provides AI-disclosure semantics tailored to the music industry. The File Identity Record connects music entities to specific digital assets. Mature provenance infrastructure provides a broader technical foundation for source, assertions and verification. Through explicit links among these layers, each can preserve its own semantics while contributing to a more complete trust record.

9. 2.0 Is Not a Conclusion: Questions We Hope to Answer with the Industry

The purpose of a Discussion Draft is to make unresolved questions explicit. The following issues will directly influence the final design of ISRC-AAM-CID 2.0. We hope to develop answers through dialogue with creators, labels, DSPs, rights organisations, standards bodies, technology platforms and legal experts, supported by further research and practical testing.

1. What is the right level of granularity for AAM as a summary of AI involvement?

Are AAM0-AAM4 still appropriate for industry use? Should percentage values be normative outputs, optional derived information, or remain only within an underlying calculation model?

2. How should dynamic weighting balance flexibility and comparability?

How should the Default AAM Profile be governed and versioned? Under what conditions and evidentiary basis should a Contextual Contribution Profile be used while preserving understandable and comparable records?

3. How far should the AI creative process be structured?

Which information should be quantified, which should use structured fields, and which should remain as narrative context?

4. How should consumer labels and industry-level disclosure be layered?

How can consumer-facing communication remain simple while labels, platforms, rights holders and other industry participants retain access to the detail they need?

5. Does the music industry need a File Identity layer?

If so, should it identify a specific file, a file version, or a higher-level digital asset? How should it relate to Recording Identity and provenance?

6. How should identity and provenance relationships persist when a file changes?

After transcoding, format conversion, loudness processing, editing or other derivative processing, how can a system express both that “this is a new digital file” and that “it still derives from the same recording or provenance chain”?

7. What Party / Representation model does the AI era require?

How should identity, authorisation and representation relationships be expressed among natural persons, Digital Representations, authorised voice models, virtual artists and fully synthetic identities?

8. How should Work-level and Recording-level disclosure relate when AAM extends beyond the recording layer?

How should Work-level and Recording-level AI involvement be recorded, calculated and displayed separately, and how should those disclosures be linked across the entities identified by ISWC and ISRC?

9. What boundaries should AI disclosure maintain from existing rights information?

How should AI disclosure, provenance, rights metadata and royalty data retain clear semantics, so transparency information can support industry use without being mistaken for mechanisms that determine copyright, ownership or revenue allocation?

10. How should AAM-CID enter the existing global standards ecosystem?

While fully reusing established standards and infrastructure, what is the most appropriate way for AAM-CID to interoperate with music data exchange, digital content provenance and related systems?

10. From Discussion Draft to Formal 2.0

ISRC-AAM-CID 2.0 will take shape progressively through industry discussion, technical research and practical validation. We intend to keep that process open so that experience from music creation, rights management, digital distribution, standards development and technology communities can inform both the design and validation of 2.0.

1. Publish the Discussion Draft

Summarise the operational experience of 1.0 and major industry developments over the past year, and set out the six directions for evolution together with the open questions that require further discussion.

2. Engage in industry discussion

Share implementation experience from 1.0 and the initial thinking behind 2.0 in appropriate industry meetings, standards environments and professional communities, and gather feedback from the music supply chain, rights management, provenance, technology and related fields.

3. Establish a public repository

Organise discussion outcomes that can be shared publicly and progressively build an issue list, terminology, data models, examples, references and version history so that the evolution of 2.0 can be continuously documented and discussed.

4. Conduct prototype validation

Test the relationships among ISWC / ISRC, AAM Records, File Identity Records and provenance information using real music files and registration workflows, and assess the usability and interoperability of the different data layers in practice.

5. Publish a 2.0 Candidate

Use industry feedback and prototype results to develop an implementable and testable candidate specification, then refine data structures, terminology and implementation approaches through public consultation.

6. Release 2.0

Once the core architecture, technical implementation, legal boundaries and interoperability direction have been sufficiently discussed and validated, publish a stable version together with implementation guidance, examples and version-management mechanisms.

From the Discussion Draft to 2.0, our aim is to develop more than a new technical specification. We want to build an open framework that matures through practice, industry discussion and collaboration.

11. Conclusion: From AI Music Identification Towards an Interoperable Global Trust Infrastructure for AI Music

In 2025, we introduced ISRC-AAM-CID to add a new dimension of identification and disclosure for AI music. More than a year of practical experience has made that direction clearer and given us a more complete understanding of the problem the framework needs to address.

The next stage is to enable different participants to understand and use a common set of clear, traceable facts around the same music entity: Which Musical Work is this? Which Sound Recording? Where did AI participate, and to what extent? Who made the relevant declaration? Which digital file does it refer to? What has happened to that file? Which information has been verified?

ISRC-AAM-CID 2.0 is evolving from a composite coding system towards a more open AI Music Identity & Trust Architecture. It builds on established identifiers such as ISWC and ISRC, uses AAM to express AI involvement, connects music entities to specific digital assets through the File Identity Record, and explores effective alignment with provenance infrastructure such as C2PA and global music data exchange ecosystems such as DDEX.

The significance of this direction is practical: AI disclosure needs to become information the music industry can actually use. Creators should be able to understand and declare it; labels, platforms and other industry participants should be able to read, exchange and verify it; and as music moves between systems, the relevant identity and context should remain clear.

What does the music industry need in order to trust AI-related music information?

That is the question at the heart of this Discussion Draft. In the next stage, we hope to work with music, technology, rights and standards communities around the world to test, discuss and refine these ideas through further practice and collaboration.

References (Public Sources)

1. Australian AI Music Alliance, “ISRC-AAM-CID: Establishing a Global Standardized AI Music Identifier” (12 February 2025). Source

2. Spotify, “Spotify Strengthens AI Protections for Artists, Songwriters, and Producers” (25 September 2025). Source

3. DDEX, “Other DDEX Initiatives – Artificial Intelligence.” Source

4. IFPI, “Music Community Introduces New Labelling Program to Distinguish Generative AI in Sound Recordings” (10 July 2026). Source

5. Coalition for Content Provenance and Authenticity (C2PA), C2PA Specifications 2.3. Source

6. Creator Assertions Working Group (CAWG), Technical Specifications. Source

7. European Commission, “Guidelines on Transparency Obligations for Providers and Deployers of AI Systems” (20 July 2026). Source

8. European Commission, “Transparency Obligations under Article 50 of the AI Act.” Source

9. Landgericht München I, Endurteil vom 31.07.2026, 42 O 763/25. Source

10. GEMA, “Court Rules in Favour of Music Creators: GEMA Prevails over SUNO” (31 July 2026). Source

Copyright, Confidentiality and Scope

The ISRC-AAM-CID framework is jointly developed and continuously advanced by the Australian AI Music Alliance and OZBeat AI. Except for third-party materials expressly identified as such, the original frameworks, models, text, diagrams and related content contained in this Discussion Draft are jointly copyrighted by the Australian AI Music Alliance and OZBeat AI. All rights are reserved.

This document is intended to support industry discussion, research and technical exploration concerning ISRC-AAM-CID 2.0 and does not constitute legal advice. References to third-party standards, frameworks, policies, judicial decisions or other public materials are provided solely to describe relevant industry developments and interoperability context. They do not imply endorsement, approval or sponsorship of ISRC-AAM-CID, AAM-CID, AAM-Verified or any view expressed in this Discussion Draft by DDEX, C2PA, IFPI, RIAA, the European Commission or any other third party.

© 2026 Australian AI Music Alliance & OZBeat AI. All rights reserved.