privacy

Musicians: Your Unreleased Work Deserves Better Cloud Storage

AI platforms are training on uploaded audio and documents. Here's how musicians and independent artists protect demos, lyrics, and unreleased recordings.

The demos that never get released are often the most honest recordings an artist makes. They’re raw, unpolished, and unguarded — which is precisely why they’re valuable as creative material, and precisely why they deserve careful handling.

Most working musicians give that material almost no thought in terms of where it lives. It ends up in Google Drive because Drive is convenient. It goes into Dropbox because the collaborator is already on Dropbox. The lyrics draft is in the Notes app because that’s what was open.

Every one of those choices is a decision about who has access to your unreleased work — and what they can do with it.

What Platforms Do With Your Audio Files

The terms of service governing cloud storage have changed materially in the past few years, largely because of the commercial value of training data for AI music generation systems.

Google’s terms of service state that the company can use content uploaded to their services to improve their products and services, including machine learning systems. The specific application to audio training data has not been confirmed by Google in public disclosures, but the terms don’t exclude it. When the terms don’t exclude it, the license doesn’t exclude it.

Adobe’s situation is a more visible case study. In 2023, Adobe updated its Creative Cloud terms in a way that appeared to grant broad rights to user content for training AI models. The immediate backlash was significant enough that Adobe issued clarifying statements — but the underlying question of what cloud storage terms actually permit has been raised and hasn’t gone away.

Dropbox updated its terms in 2023 to include language about using user content for AI features and product improvement. After user complaints, it clarified an opt-out process. The clarification is meaningful, but the default was opt-in.

The pattern across platforms is consistent: terms are being updated to permit AI training use cases, defaults are often opt-in rather than opt-out, and opt-out processes are not prominently disclosed.

The Value of What You’re Uploading

An unreleased track, a lyric draft, a voice memo with a chord progression — these aren’t just personal files. They’re intellectual property with potential commercial value.

The more specific risk isn’t that Google is reading your demos. The risk is that the content you upload enters a training pipeline that eventually contributes to a generative model, and that model produces output that is statistically derivative of content like yours without attribution or compensation.

This is the argument at the center of ongoing litigation between AI music companies and record labels. The industry-level dispute is about catalog recordings. But the same logic applies to independent artists who upload their original compositions to platforms whose terms permit AI training use.

A musician who uploads a distinctive vocal performance, a production technique, or a compositional style to a platform that uses it for model training is potentially contributing to a system that will compete with them commercially, for free, and without recourse.

Collaboration and the Shared File Problem

Collaboration tools are particularly complicated for unreleased work.

When you share a project folder with a collaborator on a file-sharing platform, you’re extending trust not just to that person but to the platform’s terms as they apply to both accounts. A collaborator using Google Drive shares files under Google’s terms. A producer sending stems through WeTransfer accepts WeTransfer’s terms, which have historically allowed them to access and share content for a range of business purposes.

Session files are especially sensitive. A DAW project file isn’t just audio — it contains the complete creative architecture of a track, including MIDI performances, plugin settings, automation data, and arrangement logic. This is the raw material of music production, and it’s typically far more revealing of technique and style than a final mix.

Sending session files over platforms not designed for sensitive professional content means accepting whatever those platforms’ terms say about content they process.

Lyrics as Protected Creative Work

Lyrics are literary works. They’re subject to the same copyright protections as any written composition, and they’re among the most commercially valuable outputs of a songwriting career.

They’re also routinely stored in the least protected ways: in Notes, in Google Docs, in messages to collaborators, sometimes on social platforms as draft posts.

The specific privacy risk of AI training on lyric drafts is worth taking seriously. Language models trained on song lyrics learn compositional patterns, rhyme schemes, thematic structures, and phrasing that are characteristic of specific writers. An AI system trained on your unpublished lyrics has, in some technical sense, access to your stylistic fingerprint before you’ve even decided whether to publish that work.

This may not cause you a concrete harm today. It represents a meaningful transfer of creative material without compensation or consent.

What Strong Protection Looks Like for Creative Files

The properties that matter most for protecting unreleased creative work are:

No AI training use. The platform’s terms and privacy policy should explicitly state that user content is not used to train AI models. “We do not sell your data” is not sufficient — training doesn’t require selling. Look for explicit language about AI model training.

Encryption that protects against internal access. Server-side AES-256 encryption means files are protected against external breaches. It also means the platform employees and systems cannot trivially read your files. This is distinct from end-to-end encryption (which would prevent even the platform from accessing content), but it meaningfully raises the barrier.

No advertising ecosystem. Platforms monetized by advertising have a structural incentive to understand user content in order to target ads. A subscription-based platform with no advertising has different incentives.

Clear data retention on deletion. When you delete files or close an account, when are they actually gone? A platform that can answer this question specifically — with a defined window and a clear policy on backup systems — is a more trustworthy custodian of professional content.

Data jurisdiction and legal exposure. Where servers are located affects what governments can demand under what legal processes. This matters for artists based outside the US who use US-based platforms, and vice versa.

A Practical System for Independent Artists

The goal is a structure that protects your creative work without creating so much friction that you abandon it within a week.

Separate personal from professional. Your personal photos and your unreleased music belong in different accounts, ideally at different services. Commingling them raises the surface area of exposure and makes access management more complicated.

Archive as you go. After every session, the final render and a labeled backup of the session file go into your private archive. This is a habit, not a project. It takes less time than exporting the mix.

Lyric and concept drafts. A single folder organized by project, with dated versions of lyric drafts, is retrievable and private. The date-labeled version history you build this way is more useful than the version history inside a collaborative document platform.

Collaboration copies vs. master copies. When you share work with a collaborator, you share a copy. The master stays in your private archive. If a collaboration ends badly, you still have your version of the work in a system you control.

Demo archive. The demos you don’t release have more potential value than most artists realize — as future reference material, as evidence of creation dates for copyright purposes, as personal history. They deserve organized storage, not the back of a hard drive that hasn’t been turned on in three years.

Private archiving of creative work has an additional practical benefit: it establishes creation dates.

Copyright in most jurisdictions exists at the moment of creation, not at registration. But establishing when something was created — if it ever needs to be proven — requires evidence. A file uploaded to a private storage service with a documented upload date is a form of contemporaneous record.

This isn’t equivalent to formal copyright registration, which remains the most robust protection and is required for statutory damages in US copyright litigation. But for independent artists who can’t register every demo, a private archive with clear dates and version history is meaningful documentation.

Daftei stores files with metadata intact, applies AES-256 encryption at rest, uses TLS 1.3 in transit, and explicitly does not use user content to train AI models. The service never sells data and never shows ads. For a working musician, that’s a materially different profile than the consumer cloud platforms that have become the default.

The 5 GB free tier covers most artists’ session needs. The Pro plan at $5.99/month or $44.99/year supports full session archives at higher storage volumes.

Your music belongs to you. The systems you store it in should reflect that.

Your memories deserve better than an ad platform.

Try daftei free →
← All posts