25 questions·answered from court dockets and official policies, dated and sourced·structured information, not legal advice
Honestly: no US court has answered this for music yet. It remains the central question in the active Suno litigation and two Sony–Udio cases as of August 19, 2026, with zero merits rulings on fair use. Meanwhile the market is answering by contract — licensing deals instead of verdicts.
A work generated entirely by AI cannot be copyrighted in the US — the human-authorship requirement was affirmed by the D.C. Circuit in March 2025, and the Supreme Court declined to review it in March 2026. A song with substantial human creative contribution can be — the AI-generated elements themselves are not protected.
Yes — synthetic artist personas are legal and increasingly common, as long as the persona doesn't imitate a real identifiable person and you meet platform disclosure rules. The risk isn't the persona; it's undisclosed AI behind claimed humanity, which platforms and press now actively expose.
Suno currently grants commercial-use rights to songs made while subscribed to Pro or Premier; Basic-plan songs are non-commercial. Check the live terms before releasing because this is a contract rule, not copyright ownership. Platforms add their own rules, and US copyright protects your human contributions rather than raw AI output.
Increasingly yes — a growing class of contests exists specifically for AI music, and mainstream opportunities are adding explicit AI rules. The answer is always in the specific contest's rules: allowed degree of AI, disclosure requirements, and voice-cloning restrictions vary per opportunity.
Cloning a recognizable artist's voice without permission is the most legally exposed thing you can do in AI music. Tennessee's ELVIS Act makes commercial voice imitation actionable, other states have right-of-publicity laws, and a federal bill (NO FAKES Act) has advanced in the Senate — though it is not law yet.
Yes — the usual DMCA machinery applies to AI tracks like any other upload, and AI adds new takedown surfaces: platforms remove impersonation and fraud-linked content on policy grounds, without any court involved. A takedown notice is an allegation, not a verdict; counter-notice procedures exist.
Partially, and the details matter. Warner exited both original cases via settlements (with prejudice against Suno, without prejudice against Udio); UMG and Capitol exited the original Udio case only, without prejudice. Sony is still litigating against both companies and filed a second Udio case in July 2026 — the disputes are not over.
Depends on the platform and on how much of the track is AI. TIDAL is explicit — wholly AI-generated tracks are ineligible for royalty attribution since July 15, 2026. Most other DSPs still pay normally unless the track trips fraud or impersonation rules. Human-made-with-AI-assistance earns royalties everywhere we track.
No general US law requires it as of August 19, 2026 — but platforms increasingly do. TIDAL labels wholly AI-generated tracks itself and removes them from royalty attribution, and the industry announced voluntary "AI-Generated" / "AI-Assisted" labels in July. Contest and distributor rules often require disclosure regardless.
The parts you wrote, edited and shaped — yes, those are protectable human authorship. Raw AI-generated lyrics pasted unchanged are in the same unprotected category as any purely machine-generated work. The more you rework, select and arrange, the stronger your claim.
Nobody outside the vendors knows. Deezer claims very high accuracy for its detector; no public test corpus, class-balance data, confusion matrix or independent replication backs any vendor's number. Every accuracy figure in circulation is a party claim, and false positives against human artists are documented anecdotally.
Keep the making. Project files, stems, version history, session recordings and drafts are the evidence that survives a detector flag, a platform dispute or a copyright registration question. No certification authority exists yet — provenance standards are being built, so for now your DAW project is your proof.
Making and releasing AI music is legal. The open legal fight is about the training side — whether AI companies may train on copyrighted recordings without a license — and no US court has ruled on that question for music yet. Separate rules govern voices and likenesses, and platforms add their own policies.
A cover of a song requires a mechanical license for the composition — same as before AI. Add a cloned voice of a recognizable artist and you stack a second, bigger problem on top: right-of-publicity law. AI vocals that don't imitate anyone real make the cover question purely a licensing one.
No. There is no such case in any US court record — zero results in CourtListener's opinion database and the RECAP federal filing archive. It appears to be a content-farm mutation of Bartz v. Anthropic, a real case about books and text models, not audio. Any article citing it as AI-music precedent is citing fiction.
They are the two track-level labels announced by the music industry coalition on July 10, 2026. AI-Generated — generative AI produced all or the primary creative elements (lead vocal, key instrumental, or the whole track from a prompt). AI-Assisted — the recording is substantially human-created, with AI used for some expressive elements.
The major labels sued Suno in June 2024 for training on their recordings. As of August 19, 2026 the case is active — Warner settled and left with prejudice, while UMG, Capitol and Sony remain in the District of Massachusetts. No fair-use or liability ruling has been made.
The original New York case is still active for Sony after UMG, Capitol and Warner exited without prejudice. The court kept Sony's DMCA count alive in April 2026, refused to add roughly 30,000 more recordings in June, and Sony responded by filing a second Udio case on July 20. No court has ruled on fair use or liability.
Tennessee's Ensuring Likeness Voice and Image Security Act — the first US state law explicitly protecting a person's voice as a property right against unauthorized AI simulation. In force since July 2024, it made commercial voice cloning actionable in Tennessee and became the template other states and the federal NO FAKES Act draw on.
The NO FAKES Act of 2026 (S.4591) would create a federal property right in a person's voice and visual likeness against unauthorized digital replicas. It was reported by the Senate Judiciary Committee on June 24, 2026 and placed on the Senate calendar — it is NOT law and has not passed either chamber.
Right now — contests, open calls, showcases and hackathons tracked on our deadlines page, each verified against official sources with its AI rules stated. The recurring flagship is the AI Song Contest; beyond contests, distributor releases, sync libraries and community showcases each have their own AI policies.
Most major generation tools allow commercial use on paid tiers and restrict it on free ones — but the specifics live in each tool's terms and change without notice. Our Tools index tracks commercial-use statements per tool with a last-verified date; treat any undated claim (including screenshots on Reddit) as expired.
Two different questions hide here. Copyright: purely AI-generated material has no US copyright owner at all (human-authorship requirement); your human contributions are yours. Contract: whether YOU may commercially use the output depends on the tool's terms of service, which vary by product and plan.
TIDAL is the clearest case — since July 15, 2026, tracks it determines are wholly AI-generated are labeled and removed from royalty attribution, and fraud-linked AI content is removed. Deezer labels AI content in playlists using its own detector. Removal generally targets fraud and impersonation, not AI use itself.
Need to act on a date? Browse live opportunities. See an error? Report a correction.