Authio Origin Classification Standard

Authio Origin Classification Standard

AOCS v1.0

A quantified technical framework for classifying Human-leading, Hybrid, and AI-leading evidence in sound recordings

Created, published and maintained by Authio / Forward Digital
Publication date: 22 July 2026
Effective date: 22 July 2026
Status: Final Authio technical standard

Abstract

Published industry initiatives distinguish between AI-Generated and AI-Assisted sound recordings. They do not define a universal numerical boundary for deciding when technical evidence in a Hybrid recording becomes AI-dominant rather than Human-dominant.

The Authio Origin Classification Standard (AOCS) defines that boundary.

Under the Authio 50% Principle, an eligible Mixed or Hybrid sound recording whose canonical, unrounded Full-Mix AI Score is greater than 0.50 is technically classified as AI-Dominant Hybrid. It MUST be routed to the AI-origin policy domain and MUST NOT be represented operationally as Human-origin.

The Full-Mix AI Score does not measure the share of creative labor, authorship, duration, stems, instruments or ownership attributable to AI. It provides a reproducible technical distinction between Human-leading and AI-leading Hybrid evidence.

AOCS separates:

1. the technical result produced by an eligible full-mix analysis;
2. the AOCS technical class;
3. the public disclosure label;
4. the legal, commercial, and economic action applied by an adopter.

This document does not claim endorsement, adoption, certification or formal compatibility by any external organization.

Document status

AOCS v1.0 is a final Authio technical standard. Authio created AOCS, administers its name, controls its normative text and publishes its official versions. Authio retains final editorial authority and decides all future modifications.

No third party has veto, co-authorship or normative authority over AOCS. Authio MAY voluntarily seek and consider external comments or establish advisory groups; participation is optional, non-binding and never a condition of validity or publication.

AOCS addresses the absence of a universal numerical boundary for classifying technical evidence in Mixed or Hybrid recordings as Human-dominant or AI-dominant.

1. Purpose

Hybrid sound recordings contain meaningful Human-oriented and AI-oriented evidence. They create a practical classification problem:

At what technical point should a hybrid recording stop being treated as Human-dominant and enter an AI-related policy workflow?

AOCS provides a deterministic technical answer based on an eligible analysis of the complete sound recording.

AOCS is designed for organizations that require a consistent classification layer, including:

• digital service providers;
• distributors and aggregators;
• labels and rights holders;
• music recognition and integrity services;
• technical and metadata providers;
• creators and creator representatives;
• standards and policy organizations.

AOCS standardizes technical classification and policy routing. It does not impose a universal legal, commercial, royalty, recommendation, or acceptance decision.

2. Scope

2.1 In scope

AOCS applies to the technical classification of a completed sound recording analyzed as a full mix.

It defines:

• the authoritative technical input used for classification;
• the Full-Mix AI Score;
• AOCS-qualified analysis requirements;
• Human-Classified, Human-Dominant Hybrid, Balanced Hybrid, AI-Dominant Hybrid and AI-Classified technical classes, and the Review Required analysis/routing status;
• the Authio 50% Principle;
• exact-boundary handling;
• binary policy-routing rules;
• treatment of invalid, abstained, and indeterminate results;
• the separation between technical classification and public disclosure;
• minimum traceability, review, appeal, governance, and versioning requirements.

2.2 Out of scope

AOCS does not determine:

• copyright ownership or legal authorship;
• consent to the use of a voice, likeness, or performance;
• whether training data, samples, compositions, or recordings were lawfully used;
• fraud, impersonation, infringement, or deceptive intent;
• royalty eligibility or payment allocation;
• recommendation, discovery, or monetization eligibility;
• whether a recording must be accepted, rejected, removed, or blocked;
• the percentage of creative labor performed by a person or a machine;
• the origin of lyrics, compositions, artwork, video, or other non-audio assets.

These matters require separate evidence, policy, contractual, and legal analysis.

3. Normative language

The terms MUST, MUST NOT, SHOULD, SHOULD NOT, and MAY are normative.

• MUST / MUST NOT indicate a requirement or prohibition for conformance.
• SHOULD / SHOULD NOT indicate a recommendation that may be departed from only for a documented reason.
• MAY indicates an optional, permitted action.

Sections explicitly marked Informative do not create conformance requirements.

4. Definitions

4.1 Exact Audio Asset

The Exact Audio Asset is the specific audio file or canonical audio payload submitted for analysis and identified by an immutable content identifier.

An AOCS class applies only to that asset.

4.2 AOCS-Qualified Analysis

An AOCS-Qualified Analysis is a completed full-mix analysis produced under an analysis profile listed in the AOCS Qualified Analysis Profile Register.

A qualified analysis MUST:

• analyze the complete eligible sound recording;
• produce an authoritative full-mix verdict;
• produce the canonical Full-Mix AI Score required by its profile;
• complete without invalidation, technical failure, insufficient audio, corrupted input, or abstention;
• preserve the analysis profile, analysis version, date, and Exact Audio Asset identifier;
• satisfy the validation, calibration, traceability, and change-control requirements of its profile;
• emit only classes for which the profile has published class-specific validation;
• return Review Required status rather than fabricate a technical class outside the profile's validated capability.

4.3 AOCS Qualified Analysis Profile

An AOCS Qualified Analysis Profile defines the approved technical conditions under which an analysis may assign an AOCS class.

Each profile MUST publicly identify:

• the profile name and version;
• the reference implementation or qualified implementation;
• the authoritative full-mix verdict used by the profile;
• the canonical Full-Mix AI Score definition;
• the canonical score representation and precision;
• the validity and abstention conditions;
• the effective date;
• the status as active, deprecated, or withdrawn;
• the applicable validation note.

Authio Forensic is the initial AOCS reference implementation maintained by Authio / Forward Digital. A future external implementation MAY be qualified only through a documented AOCS qualification process and MUST NOT claim conformance merely because it produces a numerical AI score.

4.4 Full-Mix AI Score

The Full-Mix AI Score is the calibrated technical score produced by an AOCS-Qualified Analysis of the complete sound recording.

The score is expressed from 0.00 to 1.00, where higher values indicate stronger AI-oriented evidence in the full mix. Each qualified profile MUST state its calibration target and evaluation population. The score is not an estimate of mixture share or creative contribution.

The Full-Mix AI Score:

• is the sole numerical input to the Authio 50% Principle;
• MUST be evaluated using the canonical score representation defined by the applicable qualified profile;
• MUST NOT be calculated from a rounded display percentage;
• MUST NOT be interpreted as a literal percentage of duration, stems, instruments, authorship, ownership, or creative contribution;
• MUST NOT be replaced by any subordinate diagnostic output;
• when applied to an already valid Mixed or Hybrid verdict, indicates only which side of the profile-defined technical score leads at the normative boundary; it does not measure the quantity of Human or AI contribution.

4.5 Canonical Score Value

The Canonical Score Value is the persisted or otherwise authoritative decimal score emitted by the applicable AOCS Qualified Analysis Profile at the precision declared by that profile.

The Canonical Score Value is the value of record for classification.

Display formatting MUST NOT alter the AOCS class.

4.6 Technical Origin Classification

A Technical Origin Classification is the AOCS class derived from an AOCS-Qualified Analysis.

It is a detection-based technical classification. It is not a legal or historical certification of every step in the recording's creation.

4.7 Mixed or Hybrid

A Mixed or Hybrid result indicates that the authoritative full-mix analysis contains material Human-oriented and AI-oriented evidence.

Hybrid does not mean Human-only and does not mean wholly AI-created.

4.8 Human-Dominant Hybrid

A Human-Dominant Hybrid is a valid Mixed or Hybrid result whose Canonical Score Value is below 0.50.

The recording remains hybrid and MUST NOT be represented as containing no AI involvement solely because Human-oriented evidence is dominant.

4.9 AI-Dominant Hybrid

An AI-Dominant Hybrid is a valid Mixed or Hybrid result whose Canonical Score Value is above 0.50.

The recording remains hybrid and may contain meaningful Human contribution. However, AI-oriented evidence is technically dominant.

For AOCS policy routing, AI-Dominant Hybrid:

• MUST enter the AI-origin policy domain;
• MUST NOT be represented operationally as Human-origin;
• MUST remain distinguishable from wholly AI-generated content.

4.10 Balanced Hybrid

A Balanced Hybrid is a valid Mixed or Hybrid result whose Canonical Score Value is numerically equal to 0.50 in the canonical representation declared by the applicable qualified profile.

Balanced Hybrid is a valid technical class that MUST carry Review Required status.

A value displayed as 50% does not establish that the Canonical Score Value is exactly 0.50.

4.11 Human-Classified

Human-Classified means that an AOCS-Qualified Analysis produced a valid authoritative Human full-mix verdict.

It does not prove that no generative AI tool was used at any stage of production.

4.12 AI-Classified

AI-Classified means that an AOCS-Qualified Analysis produced a valid authoritative AI full-mix verdict.

It does not claim that no person contributed prompts, performance, editing, arrangement, production, or post-processing.

4.13 AI-origin policy domain

The AI-origin policy domain is an internal or partner-facing routing category requiring an adopter to apply its AI-related disclosure, provenance, review, and policy controls.

It does not itself require rejection, removal, demonetization, non-payment, or exclusion from recommendations.

4.14 Review Required

Review Required is an analysis or routing status under which AOCS does not permit an automatic final policy-routing decision because the result is:

• Balanced Hybrid;
• invalid;
• abstained;
• incomplete;
• technically indeterminate;
• produced under a withdrawn profile;
• materially disputed by credible provenance evidence.

Where a valid technical class is materially disputed, Review Required suspends automatic routing but does not erase or replace the original technical class.

4.15 Provenance

Provenance is documented information about how a recording was created, including creator declarations, production records, project files, session evidence, source files, licenses, consent records, and relevant technical history.

4.16 Public Disclosure Label

A Public Disclosure Label is consumer-facing or partner-facing information about the use of generative AI.

It is separate from the AOCS technical class and from the adopter's commercial action.

4.17 Material Adverse Action

A Material Adverse Action is an action that may materially disadvantage a creator, supplier, recording, or rights holder, including:

• rejection from delivery;
• removal;
• suspension;
• demonetization;
• denial of royalty attribution;
• account restriction;
• fraud escalation;
• material exclusion from discovery or recommendations.

5. Core principles

5.1 Sole technical classification authority

For assigning an AOCS detection-based class, the authoritative full-mix result and Canonical Score Value are the sole technical classification inputs.

Supporting diagnostic evidence MAY explain the result but MUST NOT independently create, replace, upgrade or downgrade the AOCS class.

5.2 Technical score, not creative accounting

The Full-Mix AI Score is a calibrated technical classification score.

It is not:

• a percentage of authorship;
• a percentage of ownership;
• a percentage of Human or machine labor;
• a percentage of the recording's duration;
• a percentage of stems or instruments generated by AI.

A Canonical Score Value of 0.52 means that the eligible full-mix result is technically AI-leading under the applicable qualified profile.

It does not mean that exactly 52% of the creative work or audio was generated by AI.

5.3 Normative majority boundary

The 0.50 boundary is a normative majority boundary adopted by AOCS to distinguish the Human-oriented and AI-oriented sides of the canonical technical score for an already valid Hybrid result. “Majority” describes the score boundary, not a measured majority of duration, stems, labor, authorship, ownership, or creative contribution.

It is not presented as:

• a law of nature;
• a legal threshold;
• a copyright test;
• a measurement of creative authorship;
• proof that most creative work was performed by AI.

5.4 Classification, disclosure, and policy are separate

AOCS distinguishes:

1. the technical result;
2. the AOCS technical class;
3. the public disclosure label;
4. the adopter's legal, commercial, or economic action.

These layers MUST remain separately traceable and MUST NOT be silently collapsed.

5.5 Unknown remains unknown

Invalid, abstained, incomplete, and materially indeterminate results MUST receive Review Required status without fabrication of a technical class.

An adopter MUST NOT force a Human or AI class solely to avoid operational uncertainty.

5.6 Exact asset scope

An AOCS class applies only to the Exact Audio Asset analyzed.

A materially modified asset MUST receive a new qualified analysis before the prior class is applied to it.

Material modifications include, where they alter the analyzed signal:

• remasters;
• remixes;
• edits;
• stem mixes;
• sped-up or slowed versions;
• materially altered restorations;
• substituted masters.

5.7 No silent rewriting

A technical result MUST NOT be silently deleted, overwritten, or relabeled because a later review reaches a different operational conclusion.

The original technical result and any later review outcome MUST remain separately traceable.

5.8 No implied external endorsement

An adopter or publisher MUST NOT claim that AOCS is endorsed, adopted, certified, or approved by an external organization unless that recognition has been formally granted in writing.

6. The Authio 50% Principle

6.1 Normative rule

The Authio 50% Principle applies only when:

1. the analysis is AOCS-qualified;
2. the authoritative full-mix verdict is Mixed or Hybrid;
3. the Canonical Score Value is available;
4. the profile is active on the analysis date.

The class MUST be assigned as follows:

6.2 Display rounding

The class MUST be calculated from the Canonical Score Value, not from the rounded percentage shown in a user interface.

A score that displays as 50% may be below, equal to, or above the boundary.

6.3 Binary policy systems

When an adopter supports only binary Human-origin or AI-origin policy routing:

• AI-Dominant Hybrid MUST map to the AI-origin side;
• AI-Dominant Hybrid MUST NOT map to the Human-origin side;
• Human-Dominant Hybrid MUST remain identifiable as involving AI;
• Balanced Hybrid MUST carry Review Required status, and any result carrying Review Required status MUST NOT be forced into either side without review.

This binary-routing rule is the central operational contribution of AOCS v1.0.

6.4 Hybrid status is preserved

Dominance determines routing. It does not erase the minority contribution.

AI-Dominant Hybrid MUST remain distinguishable from AI-Classified and from wholly AI-generated content.

Human-Dominant Hybrid MUST remain distinguishable from Human-Classified.

7. AOCS technical classes and Review Required status

7.1 Review Required status

Review Required is not a sixth technical class. It suspends automatic routing.

Balanced Hybrid retains its valid Balanced Hybrid class and MUST carry Review Required status. A materially disputed valid classification retains its technical class and receives Review Required status. An invalid, abstained, incomplete or technically indeterminate analysis receives Review Required status without fabrication of a technical class. Review Required never deletes or substitutes a valid prior class.

7.2 Human-Classified is not provenance certification

Human-Classified describes the technical result.

A credible declaration of material generative use MAY require disclosure or review even when the technical result is Human-Classified.

7.3 AI-Classified is not a claim of zero Human involvement

AI-Classified describes the technical result.

It does not claim that no Human contribution occurred.

7.4 Hybrid classes retain both sides

Human-Dominant Hybrid and AI-Dominant Hybrid are both hybrid classes.

AOCS MUST NOT describe either as pure Human or pure AI.

8. Relationship to public disclosure

8.1 Separate layers

AOCS does not automatically convert every technical class into an external industry's public label.

The emerging AI-Generated and AI-Assisted labels reviewed for this standard are primarily based on how the recording was created and on which principal creative elements were generated or performed by Humans.

AOCS provides a detection-based technical classification. It does not claim to reconstruct the complete production process from audio alone.

8.2 Recommended AOCS disclosure terminology

Where a receiving system can support AOCS-specific terminology, the following terms are recommended:

8.3 Relationship to AI-Generated and AI-Assisted labels

A final AI-Generated or AI-Assisted public label SHOULD incorporate reliable provenance when available.

An AOCS class MUST NOT be represented as legal proof of the complete creative process.

However:

• AI-Dominant Hybrid MUST NOT be represented as Human-only;
• AI-Dominant Hybrid MUST enter the AI-origin policy domain;
• when provenance is unavailable, AI-Dominant Hybrid SHOULD receive an explicit AI-related disclosure;
• no compatibility with an external label scheme may be claimed without formal agreement.

8.4 Binary public-disclosure fallback

When a receiving system supports only a binary public distinction between Human-origin and AI-origin, AI-Dominant Hybrid MUST be placed on the AI-origin side.

The disclosure SHOULD make clear that the result is hybrid and technically AI-dominant rather than wholly AI-generated.

9. Supporting evidence

9.1 Supporting diagnostics

Supporting diagnostic outputs MAY explain the authoritative full-mix result.

They MAY provide additional context for a documented review.

They MAY assist interpretation within a separately validated capability.

Supporting diagnostics are subordinate to the authoritative full-mix result. They MUST NOT be represented as independent calibrated probabilities unless a separately validated capability supports that claim.

They MUST NOT independently create an AOCS class.

They MUST NOT replace the Full-Mix AI Score.

They MUST NOT independently upgrade or downgrade the AOCS class.

Unsupported, unstable or abstained diagnostics MUST NOT be asserted as fact.

10. Provenance, declaration, and conflicting evidence

10.1 Complementary evidence

Technical detection and provenance are complementary.

Creators, labels, distributors, and suppliers SHOULD disclose material generative use and SHOULD retain supporting production evidence.

10.2 Conflict handling

When credible provenance materially conflicts with the technical classification, the recording MUST enter a documented review.

The review MAY:

• confirm the AOCS class;
• require a new analysis of the correct asset;
• invalidate the original analysis;
• apply a different public disclosure based on verified provenance;
• assign Review Required status;
• supersede the operational classification with a reasoned decision.

The original technical result MUST remain preserved.

10.3 No automatic downward override

A self-declaration alone SHOULD NOT automatically downgrade AI-Dominant Hybrid or AI-Classified to Human-origin.

A contrary outcome requires sufficient evidence and a documented review.

11. Boundary review and uncertainty

AOCS v1.0 defines a deterministic technical boundary at 0.50.

It does not define a universal uncertainty band around that boundary.

An adopter MAY require additional review before applying a Material Adverse Action to a result near 0.50.

Any adopter-specific review band MUST:

• be documented;
• be applied consistently;
• preserve the original AOCS class;
• remain separate from the Authio 50% Principle;
• not be presented as a different AOCS threshold.

A future universal AOCS review band MUST be supported by published validation and introduced through a versioned amendment.

12. Review and appeal

12.1 Mandatory review for adverse actions

An adopter MUST provide a documented review and appeal process when an AOCS result is used to apply a Material Adverse Action.

The process MUST:

• identify the Exact Audio Asset analyzed;
• identify the qualified profile and analysis version;
• disclose the AOCS class used for the action;
• allow submission of relevant provenance evidence;
• provide a reasoned outcome;
• distinguish the original technical result from the review decision;
• retain an auditable history.

12.2 Review for non-adverse uses

For informational, research, or internal-routing uses that do not create a Material Adverse Action, an adopter SHOULD provide a reasonable correction process.

12.3 Evidence considered in review

A review MAY consider:

• the exact asset submitted for analysis;
• the analysis date and qualified profile;
• the Canonical Score Value and authoritative verdict;
• production explanations;
• project files, session exports, or stems;
• licenses, consent records, and provenance documents;
• evidence that the analyzed asset differs from the released asset;
• evidence that the analysis profile was invalid, deprecated, or inapplicable.

12.4 Review outcomes

A review outcome SHOULD identify whether it affects:

• the technical result;
• the AOCS class;
• the public disclosure;
• the adopter's policy action.

These are distinct records.

13. Traceability and versioning

An implementation claiming AOCS v1.0 conformance MUST retain:

• the AOCS version;
• the AOCS Qualified Analysis Profile;
• the analysis implementation and version;
• the analysis date and time;
• the immutable identifier of the Exact Audio Asset;
• the authoritative full-mix verdict;
• the Canonical Score Value;
• the AOCS technical class;
• the validity, abstention, or invalidation status;
• any later review status and reasoned outcome.

The exact storage format is implementation-specific.

AOCS does not require disclosure of proprietary implementation details, parameters or internal decision logic.

13.1 Reanalysis

A later analysis version MAY produce a different result.

Reanalysis MUST create a new versioned classification record and MUST NOT silently rewrite the prior record.

13.2 Profile withdrawal

A withdrawn profile MUST NOT be used for new AOCS classifications after its withdrawal date.

Existing classifications produced while the profile was active MUST remain historically traceable and MAY require review or reanalysis according to the published withdrawal notice.

13.3 Score comparability

A qualified profile MUST NOT claim score comparability with another profile or version unless that comparability is supported by the applicable validation documentation.

14. AOCS qualification and reference implementation

14.1 Initial reference implementation

Authio Forensic is the initial AOCS reference implementation maintained by Authio / Forward Digital.

AOCS-AUTHIO-FORENSIC-1.0 is the active Authio reference profile published with this standard. Its evidence basis, scope and limitations are defined in the accompanying register and validation note.

14.2 External implementations

An external system MAY become AOCS-qualified only if it:

• satisfies the published technical and validation requirements;
• exposes the required canonical score semantics;
• supports the required validity and abstention states;
• preserves traceability;
• passes the applicable qualification evaluation;
• receives a reasoned qualification decision for a named profile and version under published criteria.

Qualification decisions MUST identify the evidence considered and unmet requirements, MUST manage conflicts of interest and reviewer recusals, and MUST be appealable. Authio MUST NOT deny qualification solely because an implementation competes with Authio.

Producing a score between 0.00 and 1.00 is not sufficient for AOCS qualification.

14.3 Qualification register

Authio MUST maintain a public register containing:

• active qualified profiles;
• reference and external implementations;
• effective dates;
• profile versions;
• applicable validation notes;
• deprecated and withdrawn profiles;
• qualification and withdrawal decisions.

14.4 No implied certification

AOCS conformance does not imply that Authio has audited an adopter's complete business, legal, or policy process unless a separate certification program exists.

15. Conformance

15.1 Classification conformance

An implementation claiming AOCS Classification Conformance MUST:

1. use an AOCS-Qualified Analysis;
2. use the Canonical Score Value;
3. apply the Authio 50% Principle only to Mixed or Hybrid results;
4. classify a score above 0.50 as AI-Dominant Hybrid;
5. classify a score below 0.50 as Human-Dominant Hybrid;
6. classify a score equal to 0.50 as Balanced Hybrid;
7. assign Review Required status to invalid, abstained, incomplete, and indeterminate outcomes without fabricating a technical class;
8. keep all supporting diagnostic evidence subordinate to the authoritative full-mix result;
9. preserve asset and version traceability.

15.2 Policy-routing conformance

An adopter claiming AOCS Policy-Routing Conformance MUST additionally:

1. route AI-Dominant Hybrid to the AI-origin policy domain;
2. never represent AI-Dominant Hybrid as Human-origin;
3. preserve Human-Dominant Hybrid as a hybrid class with AI involvement;
4. keep technical class, disclosure, and policy action separate;
5. provide the mandatory review and appeal process for Material Adverse Actions;
6. avoid claiming external endorsement without authorization.

15.3 Permitted conformance statement

When using an active qualified profile, an adopter MAY publish:

This organization follows the Authio Origin Classification Standard v1.0. Eligible Mixed or Hybrid recordings with a Canonical Full-Mix AI Score above 0.50 are classified as AI-Dominant Hybrid and routed to the organization's AI-origin policy domain.

The statement MUST identify whether the adopter claims Classification Conformance, Policy-Routing Conformance, or both.

16. Worked examples

16.1 AI-leading hybrid

Authoritative result: Mixed or Hybrid
Canonical Full-Mix AI Score: 0.52
AOCS technical class: AI-Dominant Hybrid
Required routing: AI-origin policy domain
Recommended disclosure: AI-Dominant Hybrid: Technical Classification

The recording contains Human-oriented and AI-oriented evidence, but AI-oriented evidence leads in the full mix.

It MUST NOT be represented as Human-origin.

This does not claim that exactly 52% of the creative work or audio was generated by AI.

16.2 Human-leading hybrid

Authoritative result: Mixed or Hybrid
Canonical Full-Mix AI Score: 0.48
AOCS technical class: Human-Dominant Hybrid
Required routing: Hybrid workflow with Human-oriented dominance
Recommended disclosure: AI-Involved: Human-Dominant Hybrid

The recording remains hybrid and MUST NOT be represented as containing no AI involvement.

16.3 Rounded display at 50%, canonical score above the boundary

Authoritative result: Mixed or Hybrid
Canonical Full-Mix AI Score: 0.504
Displayed score: 50% AI
AOCS technical class: AI-Dominant Hybrid

The display percentage does not determine the class.

16.4 Exact boundary

Authoritative result: Mixed or Hybrid
Canonical Full-Mix AI Score: numerically equal to 0.50 under the applicable profile
AOCS technical class: Balanced Hybrid
Required routing: Review Required status

16.5 AI-oriented full mix

Authoritative result: AI
AOCS technical class: AI-Classified
Required routing: AI-origin policy domain

The class does not claim zero Human involvement.

16.6 Invalid or abstained analysis

Authoritative result: Invalid, abstained, or technically indeterminate
AOCS technical class: none fabricated Analysis/routing status: Review Required

No Human or AI class may be fabricated.

17. Required Technical Basis and Validation Note

Authio publishes the AOCS Technical Basis and Validation Note v1.0 as a companion to this standard.

The note MUST provide sufficient non-proprietary information to evaluate the framework, including:

• the public definition of the Full-Mix AI Score;
• the initial qualified analysis profile;
• the canonical score representation and precision;
• the public calibration target and the basis used to assess calibration;
• the scope and composition of the evaluation program;
• the nature of the evidence supporting the profile's published claims;
• known limitations;
• available evidence and known limitations concerning boundary behavior near 0.50;
• false-positive and false-negative considerations;
• abstention and invalidation conditions;
• versioning and reproducibility controls;
• change-control requirements;
• score-comparability limits across versions.

The validation note MUST NOT expose proprietary implementation details, private file paths, credentials or operational secrets.

The 0.50 boundary is a normative majority boundary. The validation note is required to state the evidence scope, profile conditions and known limitations relevant to decisions around that boundary.

18. Governance

AOCS is created, defined, published, controlled and maintained by Authio. Authio retains sole normative and editorial authority, administers the AOCS name, controls official publication and decides all future modifications.

Authio MAY request and consider external feedback and MAY create advisory groups. Such participation is voluntary and non-binding. It does not transfer authority, authorship or maintenance rights; create a third-party veto; or condition the validity of an AOCS version.

Authio will document material changes and publish versioned changelogs. Authio alone decides whether feedback is accepted and whether a new version, amendment, deprecation or withdrawal is issued.

Material changes to any of the following require a new published version:

• the 0.50 boundary;
• technical class definitions;
• policy-routing requirements;
• Canonical Score Value semantics;
• qualification requirements;
• mandatory review protections.

Editorial corrections that do not change normative meaning MAY be published as minor revisions.

Authio MUST maintain:

• a public version history;
• a material change log;
• a deprecation and withdrawal policy;
• effective dates;
• archived superseded versions;
• a public record of active qualified profiles.

19. Relationship to metadata standards

AOCS defines classification semantics. It is not a replacement for music-industry metadata standards.

DDEX's ERN data model supports communicating whether a SoundRecording contains generative-AI contribution and whether contribution is none, partial, or all.

AOCS MAY inform future metadata mappings, but this standard does not claim formal DDEX adoption, compatibility, certification, or endorsement.

Any future mapping SHOULD preserve:

• the AOCS technical class;
• the Canonical Full-Mix AI Score where exchange is appropriate;
• the AOCS version;
• the qualified analysis profile and version;
• the distinction between technical classification and creator declaration;
• the distinction between AI-Dominant Hybrid and wholly AI-generated content;
• the review status.

A future implementation profile SHOULD be developed separately with relevant standards bodies and supply-chain participants.

20. Informative industry context

This section is informative, time-sensitive, and based on the cited sources as accessed on 21 July 2026. It does not establish endorsement or formal compatibility.

In July 2026, a coalition including IFPI, RIAA, A2IM, WIN, IMPALA, The Grammys, SAG-AFTRA, and the Human Artistry Campaign announced voluntary track-level labels distinguishing AI-Generated and AI-Assisted sound recordings.

Those labels focus on the role of generative AI in the creative process.

TIDAL's published policy currently focuses action on music it determines to be wholly AI-generated. TIDAL states that it may expand its approach to substantially AI-generated content as detection technology becomes more reliable.

DDEX's ERN data model includes fields for communicating whether a SoundRecording contains content generated using AI and whether AI contribution is partial or complete.

The published initiatives reviewed for this standard do not specify a universal numerical threshold for classifying technically Mixed or Hybrid recordings as Human-dominant or AI-dominant.

The Authio 50% Principle addresses that specific gap within AOCS.

21. Conclusion

AOCS v1.0 establishes one narrow, clear, and reproducible principle:

An AOCS-qualified Mixed or Hybrid sound recording with a Canonical Full-Mix AI Score greater than 0.50 is AI-Dominant Hybrid, MUST enter the AI-origin policy domain, and MUST NOT be represented operationally as Human-origin.

The standard preserves six essential distinctions:

1. Hybrid is not the same as wholly AI-created.
2. AI-Dominant Hybrid is not Human-origin for technical policy routing.
3. Human-Dominant Hybrid still contains AI involvement.
4. The Full-Mix AI Score is not a percentage of authorship or creative labor.
5. Technical classification is not a legal or commercial judgment.
6. Public disclosure may incorporate provenance in addition to technical detection.

Authio publishes and maintains AOCS v1.0 as its official technical standard for a quantified boundary in hybrid sound recordings.

References

1. IFPI, Music community introduces new labelling program to distinguish generative AI in sound recordings, 10 July 2026.
   https://www.ifpi.org/music-community-introduces-new-labelling-program-to-distinguish-generative-ai-in-sound-recordings/

2. RIAA, Music Community Introduces New Labeling Program To Distinguish Generative AI in Sound Recordings, 10 July 2026.
   https://www.riaa.com/music-community-introduces-new-labeling-programto-distinguish-generative-ai-in-sound-recordings/

3. TIDAL Support, AI Policy, updated 20 July 2026.
   https://support.tidal.com/hc/en-us/articles/48031883413521-AI-Policy

4. DDEX Data Dictionary, ContainsAI, ERN Edition.
   https://service.ddex.net/dd/DD-ERN-432/dd/ddex_ContainsAI.html

5. DDEX, Release Delivery: Electronic Release Notification Message Suite.

   https://ddex.net/standards-2023/release-delivery/

End of AOCS v1.0. Status: Final Authio technical standard.