Adversarial Audio Protection for Artist-Owned Music Catalogues
MusicPoison is a service designed to help artists protect their music from unauthorized AI training while preserving a legitimate licensing path for companies that want authorized clean audio access.
A dual approach to audio distribution — protection by default, clean access by authorization.
For unlicensed distribution, MusicPoison applies inaudible adversarial perturbations intended to disrupt AI training feature extraction while preserving normal human listening quality.
For authorized buyers, MusicPoison supports clean audio distribution with embedded cryptographic proof-of-license through inaudible watermarking.
Three core modules powering the MusicPoison protection pipeline.
5-layer adversarial audio protection system.
Track registration, signed license keys, registry logic, and embedded watermark proof.
Verification layer for perceptual quality, AI-feature drift, robustness, and watermark detection.
Artist prepares final master audio.
MusicPoison creates a protected distribution copy.
Unlicensed copies remain protected by default.
Authorized buyers request a license.
Licensed clean copies are delivered with embedded proof-of-license.
Verification tools can inspect watermark provenance.
The evaluation tier grants documentation access for technical review. Full capabilities are available under an enterprise licensing agreement.
Enterprise terms are negotiated directly. Contact for licensing discussion.
For enterprise licensing discussions, partnership proposals, or integration enquiries — reach out directly.
Name
Antonios Valamontes
Organization
Kapodistrian Academy of Science
Project
MusicPoison SDK
Purpose
Artist IP Protection and Authorized AI Training Licensing
Access the MusicPoison SDK Developer Documentation v1.0 — an early-stage technical reference covering the core architecture, protection layers, and licensing model. Provided as-is for evaluation purposes.