AI Audio Protection Standard — Public Draft 1.0
Active StandardAIAPS
A standard for declaring and verifying audio recordings not authorized for AI use.
The AI Audio Protection Standard (AIAPS) defines a framework for identifying, marking, and verifying audio recordings whose rights holders do not authorize use in AI training, dataset ingestion, voice cloning, style modeling, or synthetic generation without permission.
AIAPS specifies the use of audio fingerprinting, embedded watermarking, and registry-backed records to establish a consistent and verifiable identity for protected recordings.
Section 01 — Purpose
Purpose of the Standard
AIAPS establishes a standardized method for rights holders to signal that an audio recording is not authorized for use in AI systems without permission.
The standard defines how such declarations are encoded, registered, and verified in a consistent and interoperable manner.
Section 02 — Protection Methods
Technical Components
An AIAPS-compliant recording incorporates the following components.
01
Audio Fingerprint
A deterministic identifier derived from the spectral characteristics of the recording, enabling identification across transformations such as compression, normalization, and format conversion.
02
Embedded Watermark
A machine-detectable signal encoded within the audio waveform, carrying structured data associated with the AIAPS designation.
03
Registry Record
A timestamped record linking the fingerprint, watermark payload, and rights-holder declaration to a unique AIAPS identifier.
Section 03 — Conformance
AIAPS-Conformant Recordings
A recording is considered AIAPS-conformant if it includes all required components.
Section 04 — Declaration
Rights-Holder Declaration
AIAPS Protection Certificate
VerifiedAIAPS Protected — NO_AI_TRAINING
Usage Policy
This recording is registered under the AI Audio Protection Standard (AIAPS) and is designated as not authorized for AI training, dataset ingestion, voice cloning, style modeling, or synthetic generation without permission from the rights holder.
AIAPS Registry — Verified Record
Rev. 1.0
Each AIAPS-protected recording carries a rights-holder declaration indicating that the recording is not authorized for AI-related use without permission.
This declaration is represented within the watermark payload, associated registry record, and supporting metadata where applicable.
Section 05 — Workflow
Reference Implementation Workflow
The following describes the standard implementation process.
Step 01
Fingerprint Generation
A fingerprint is computed from the input recording using a deterministic algorithm defined by the implementation.
Step 02
Watermark Encoding
A structured payload containing the AIAPS identifier and policy declaration is encoded into the audio signal.
Step 03
Registry Submission
The recording is registered with a timestamped entry linking all associated components.
Step 04
Verification
The recording can be analyzed to extract identifying signals and matched against the registry to confirm conformance and declared policy.
Section 06 — Implementation
Reference Plugin
The AIAPS Reference Plugin provides an implementation of the standard for use within digital audio workstations.
Processing is performed locally. Audio data does not leave the user's environment during fingerprinting or watermark encoding.
Availability
Reference Implementation Availability
The AIAPS Reference Plugin will be made available to creators to enable conformance with the standard. Registration for early access is currently open.