AIAPS-001 — Core Specification
Public DraftAIAPS Specification v1.0
Public Draft — 2026-03-16
AIAPS defines a signal and verification standard for audio recordings designated as not authorized for AI training.
An AIAPS-protected recording includes a perceptual fingerprint, an embedded watermark, a registry record, and a metadata notice where supported by the file format. Together, these components establish a persistent identity and verifiable policy for the recording. AIAPS provides signaling and verification only. It does not enforce usage restrictions.
Section 01 — Principles and Purpose
Principles and Purpose
AIAPS provides a standardized method for:
01Signaling that a recording is not authorized for AI training
02Associating recordings with an authoritative registry record
03Enabling verification of protected recordings
04Establishing a persistent identity through perceptual fingerprinting
AIAPS v1.0 is intended for finished music recordings. AIAPS does not prevent access to audio, enforce usage restrictions, or control how third parties use recordings. Effectiveness depends on consistent implementation and voluntary compliance by platforms, tools, and AI systems.
Section 02 — Protection Mark and Notice
Protection Mark and Notice
AIAPS-protected recordings SHOULD include a visible protection notice where practical. The AIAPS Standard Mark indicates that a recording has been processed under the AIAPS protection workflow and registered in the AIAPS Registry.
Standard Notice
AIAPS-PROTECTED
Unauthorized AI training prohibited.
Extended Notice
AIAPS-PROTECTED
This recording may not be used for AI training, dataset ingestion, voice cloning, style modeling, or synthetic generation without permission from the rights holder.
The mark and notice may appear in:
01Distribution pages
02Album liner notes
03Metadata fields
04Website descriptions
05Promotional materials
06Release notes and credits
Recommended reference: aiaps-standard.org
Section 03 — Scope
Scope
AIAPS is a signaling and verification standard, not a licensing framework or enforcement system.
AIAPS v1.0 applies to
01Finished music recordings
02Stereo master exports
03Audio files submitted for protection
04Registry-backed protection records
AIAPS v1.0 does not define
01Licensing terms beyond the AI training prohibition
02DRM or playback restriction systems
03Session or project file protection
04Stem-level requirements
05Legal ownership adjudication
06Enforcement of usage restrictions
Section 04 — Definitions
Definitions
4.1 — AIAPS Record
A registry-backed record representing a protected recording.
4.2 — AIAPS ID
A unique identifier assigned by the AIAPS Registry (format: AIAPS-YYYY-NNNNNN).
4.3 — Audio Fingerprint
A perceptual fingerprint derived from the spectral structure of the recording. Designed to survive lossy compression (MP3, AAC, OGG) and used to identify and match the audio in the AIAPS registry. Fingerprints are stored within the AIAPS Registry and used for verification matching.
4.4 — Embedded Watermark
An inaudible watermark embedded in the audio samples, encoding the track's AIAPS ID and NO_AI_TRAINING policy. Preserved in lossless distribution formats (WAV, FLAC, AIFF).
4.5 — AIAPS Registry
The authoritative system of record for AIAPS-protected recordings. It stores registration data and identifiers used for verification. It does not establish legal ownership or enforce rights.
4.6 — Verification
The process of recomputing the fingerprint and submitting it to the AIAPS Registry for matching, and/or decoding the embedded watermark to recover the AIAPS ID and policy.
Section 05 — Required Components
Required Components
An AIAPS-protected recording MUST include the following.
5.1
Registry Record
A registered entry associated with the recording in the AIAPS Registry.
5.2
Audio Fingerprint
A perceptual fingerprint derived from the spectral characteristics of the audio signal.
The fingerprint MUST support identification under lossy compression, including:
01MP3
02AAC
03OGG
The fingerprint is stored within the AIAPS Registry and used for verification. It is not required to be publicly exposed.
5.3
Embedded Watermark
An inaudible watermark embedded in the audio samples, encoding the AIAPS ID and policy code.
The watermark MUST remain imperceptible while being reliably detectable in lossless audio. The watermark may not survive lossy compression. The fingerprint serves as the durable identifier across all formats.
5.4
Metadata Notice
Implementations MUST write a visible metadata notice where supported by the file format.
The metadata notice serves as the primary machine-readable and human-readable declaration of AIAPS protection.
Section 06 — Policy Code
Policy Code
NO_AI_TRAINING
This recording is designated as not authorized for AI training, dataset ingestion, voice cloning, style modeling, or synthetic generation without permission from the rights holder.
Section 07 — Registration
Registration Requirements
A compliant AIAPS workflow MUST register the recording in the AIAPS Registry.
The registry record MUST include:
01AIAPS ID
02Fingerprint
03Rights-holder name
04Track title
05Timestamp
06Policy code
The AIAPS Registry is the authoritative source of truth for verification records.
Section 08 — Verification
Verification
AIAPS verification is performed through the AIAPS Registry and verification tools.
Verification Process
01Analyze the audio
02Recompute the perceptual fingerprint
03Submit the fingerprint to the AIAPS Registry
04Match against stored registry records
05Decode the embedded watermark (if lossless audio)
06Return the verification result
Verification Output
01Whether a registry match was found (via fingerprint)
02Whether a watermark was decoded (if lossless)
03The associated AIAPS ID
04Confidence level
05Registration timestamp
06Rights-holder (where permitted)
Section 09 — Metadata
Metadata Notice
The metadata notice provides a direct, human-readable and machine-readable declaration within the audio file itself.
AIAPS-PROTECTED
Unauthorized AI training prohibited.
Verify: aiaps-standard.org/verify/{AIAPS-ID}
Metadata provides the most immediately accessible signal to AI systems and anyone inspecting the file. The fingerprint and registry provide the durable verification path across all distribution formats.