MusicShield Beta
To protect your music during the Beta, you'll need a free MusicShield Key. Click Get MusicShield Key, complete the short form, and we'll send your key straight to your inbox. Each key lets you protect up to 5 tracks.
Upload & Protect
Drop audio files or browse
Supported during Beta: MP3 and WAV files, up to 25 MB, sample rate 16 kHz - 48 kHz
Frequently Asked Questions
What kinds of audio files are supported in the beta?
During the beta, you can upload MP3 or WAV files up to 25 MB with sample rates between 16 kHz and 48 kHz. We're actively expanding compatibility and will support more formats in the stable release.
What is the output format in the beta, and how long does processing take?
During beta, the protected output is a 320 kbps MP3. Processing may take longer while we scale; we expect significantly faster turnaround in the stable release.
Do you store my uploads?
We don’t retain your audio. We process your file to generate a protected copy, then delete both the upload and the protected file within approximately one hour of processing. See our Terms of Service for details.
Do you train models on my music?
No. We never use your uploads - or any data derived from them - to train, fine-tune, or improve models.
What metadata do you keep?
We may retain de-identified operational metadata (e.g., file size, checksum, processing time) for security and reliability analytics — never the audio itself.
Will protection change how my track sounds?
MusicShield is designed to be perceptually transparent. It applies psychoacoustically-masked micro-signals that sit below hearing thresholds and within regions already masked by the music.
Where can I report bugs or request features?
You can reach out through our Contact page or join our Discord community to report bugs, request features, or get help from other beta users.
How can I test the effectiveness of MusicShield?
We recommend starting with the music examples on the MusicShield page. You can listen to protected vs. unprotected samples and their corresponding model generations. Once available, check our forthcoming paper for full details.
Verifying protection against unauthorized model training is non-trivial because it requires training or fine-tuning an AI model on your tracks from scratch. In our evaluations, we tested a broad set of state-of-the-art open-source systems—including MusicGen, MusicLM, JASCO, and Riffusion—and demonstrate superior effectiveness against unauthorized AI training.
For a quick editing check, you can upload a MusicShield-protected clip to models with simple UIs (e.g., MusicGen or Producer.ai), apply typical melody-conditioned AI editing, and compare results to the same edits on an unprotected clip. You should observe that protected audio resists faithful editing or yields noticeably degraded/less on-target edits compared to the unprotected version.
We’ve included a short video below that shows this process step-by-step. A more detailed evaluation-testing module will be available soon to help you easily visualize the protection level for your own tracks.
Is MusicShield future-proof?
No security is permanent—AI tools evolve. MusicShield is designed to evolve too: we version the shield, continuously test against new training and editing pipelines, and ship updates. If a new model weakens protection, simply re-shield your track with the latest version to restore strength (your original audio stays unchanged). Effectiveness can vary by model and task, so for high-stakes releases we recommend defense-in-depth (e.g., keep masters private, use provenance tools). Our paper (releasing soon) details evaluations, and we welcome bypass reports to harden future updates.
How can I help support MusicShield?
The best help is real-world use and honest feedback—try the beta, share your results, and tell a friend who might benefit. If you can connect us with artists, labels, or platforms—or want to collaborate on research, pilots, or investment—drop us a note. If you’d like to chip in, you can also make an optional contribution below. Contributions help us keep evolving MusicShield to stay ahead of rapidly evolving AI models—so we can better protect artists’ work, ship improvements faster, reduce processing time, and expand format/support coverage.