All articles

Someone Released My AI Song Before Me. What You Can (and Can't) Do

It happened to me in 2024 and it still stings. I shared a rough Suno track in a Discord server on a Tuesday, and by Saturday the exact same audio was sitting on Spotify under a stranger’s artist name, complete with cover art and a release date that beat mine by a week. If you’re reading this because someone released your AI song before you did, take a breath. You have more options than the doom posts on Reddit suggest, but fewer than you’d have with a fully human recording, and the difference matters enormously.

Here’s the short version before we go deep. A purely AI-generated song almost certainly has no copyright protection in the United States, which means you cannot sue the thief for copyright infringement on the audio itself. But you can still get the stolen upload taken down in many cases, you can claim copyright on any human elements like lyrics you wrote, you can use platform impersonation and mismatch policies that didn’t exist two years ago, and you can build an evidence trail that makes you nearly impossible to rob a second time. The rest of this article walks through each of those, with the real deadlines, the real forms, and the places where I think the common advice is flat wrong.

Why This Keeps Happening to AI Musicians

The scale of this problem surprised even me when I dug into the numbers. Deezer reported receiving over 30,000 fully AI-generated tracks every single day by late 2025, up from 10,000 in January of that year, according to reporting from Music Business Worldwide. When that much synthetic audio floods into distribution pipelines, opportunists follow. Raw AI audio files carry no embedded project data, no stems, no session files, nothing that ties them to you. Whoever uploads first to a distributor often becomes the de facto owner in the eyes of every automated system downstream.

The theft usually happens at the sharing stage. You post a track to Discord, Reddit, TikTok, or the public feed inside Suno or Udio itself to get feedback. Someone rips the MP3, runs it through a cheap distributor, and it’s live on streaming platforms within days. Some thieves speed the audio up or down slightly to dodge duplicate detection. Philadelphia musician Ben Lyman-Schmidt, who records as Makeshift Hammer, documented exactly this scheme in a widely shared piece for Philadelphia Magazine in July 2026, describing fake artists siphoning his royalties with speed-altered copies and bot listeners inflating the streams. His songs were human made, but the playbook is identical for AI tracks, and AI tracks are softer targets.

So the first uncomfortable truth: this isn’t a freak event that happened to you. It’s an industrial process, and your song was inventory.

Now for the part that determines everything else. Under current US law, a song generated entirely by AI from a text prompt is not protected by copyright, because copyright requires a human author.

This isn’t a gray area anymore. The US Copyright Office released Part 2 of its report on copyright and artificial intelligence on January 29, 2025, after reviewing more than 10,000 public comments, and its conclusion was blunt. Human authorship is the bedrock of copyrightability, and works generated entirely by AI cannot be registered. Crucially, the Office also found that prompts alone do not provide enough creative control to count as authorship, no matter how detailed or clever those prompts were. You can spend three hours refining a prompt and the resulting audio still belongs to no one under copyright law.

The courts backed this up. On March 21, 2025, the US Court of Appeals for the DC Circuit ruled in the long-running Thaler case that the Copyright Act requires a human author, and the court denied a rehearing on May 12, 2025, according to the Congressional Research Service. Congress could change this someday. It hasn’t, and the Copyright Office explicitly recommended against creating new rights for AI outputs.

What does this mean when someone steals your track? It means the thing you’d instinctively reach for, a copyright infringement claim on the recording, probably doesn’t exist. As the music rights platform Orphiq puts it in their guide for artists, the practical risk of a fully AI-generated song is that anyone can use it without licensing it and you have no legal mechanism to stop them under copyright law. The song sits in something close to the public domain from the moment it’s generated.

I want to be precise here, because people overcorrect in both directions. The audio itself, if the AI made all of it, is unprotected. But that is not the same as saying you have zero rights or zero moves. Keep reading.

What You Actually Own When You Use Suno or Udio

Your relationship with the AI platform is a contract, and contracts give you something even when copyright gives you nothing. The details changed a lot between 2024 and 2026, so let me lay out where things stand.

Suno currently runs three tiers: Free at $0, Pro at $10 a month, and Premier at $30 a month, with annual billing bringing those down to an effective $8 and $24 respectively. The dividing line that matters is commercial rights. On the free tier you own nothing. Free-tier songs are for personal, non-commercial use, Suno retains the rights, and under the 2026 terms free users are losing audio downloads entirely. Subscribing later does not retroactively grant you commercial rights to songs you made while free, a trap that catches people constantly. Songs generated while you held a paid plan, on the other hand, keep their commercial status even if you later cancel, as the distribution analytics firm Dynamoi documents in its breakdown of Suno’s commercial rights by tier.

But notice the vocabulary shift. After Warner Music Group settled its copyright lawsuit with Suno in November 2025, and Universal settled with Udio around the same period, the platforms rewrote their terms. Suno’s language moved away from the word ownership toward granted commercial rights, and the terms now explicitly warn that no copyright is guaranteed to vest in your output. The legal analysts at terms.law summarize it well: your protection comes from Suno’s contractual terms, not traditional copyright law, and Suno does not indemnify you if anything goes wrong. Udio users got an even ruder awakening; after its Universal settlement the company gave users a window of just 48 hours to download their existing songs before shifting to a new walled-garden model, as the Associated Press reported.

Here’s the catch that most articles skip, and it’s the one that matters for your situation. The thief who released your song never signed Suno’s terms of service regarding your track. Your contract is with Suno, not with them. A license from Suno gives you permission to exploit the track commercially; it does not conjure an enforceable property right against third parties that copyright law says doesn’t exist. Some Suno users argue loudly online that their subscription makes them the copyright holder, and I’ve watched those arguments collapse the moment a lawyer looks at them. The subscription buys you a commercial license and a clean position with distributors and platforms. It does not buy you a copyright lawsuit.

So Is There Anything You Can Actually Do

Yes, and this is where I disagree with the fatalists. The absence of a slam-dunk copyright claim does not mean the absence of leverage. In practice, you have four workable fronts: the DMCA process through platforms and distributors, streaming platform policies on impersonation and content mismatch, copyright claims on the human-authored pieces of your song, and prevention infrastructure for everything you release from now on. Most stolen-track situations get resolved through the first two, without a courtroom ever entering the picture.

Let me take them in order of speed.

The DMCA Route and How to Run It Properly

The Digital Millennium Copyright Act, the US law that lets rights holders demand removal of infringing content from online platforms, is the standard weapon, and it works faster than people expect even in murky AI cases. A takedown notice goes to the platform hosting the stolen upload, the platform removes the content to keep its own legal safe harbor, and the burden shifts to the uploader to fight back.

Now, the honest caveat. A DMCA notice requires you to state, under penalty of perjury, that you own or control the copyrighted work. If your song is 100 percent AI-generated with zero human elements, a pure DMCA claim on the audio stands on shaky ground, and a sophisticated thief could call your bluff. In the real world, almost nobody calls the bluff, because the uploader would have to file a counter-notice containing their real name and address and consent to federal court jurisdiction. Thieves running fake-artist schemes at scale don’t want their identity in a legal filing. The Florida entertainment law firm BrewerLong notes in its guidance on AI music that platforms may remove flagged tracks immediately and that even wrongful takedowns force the other side into the counter-notice process to get anything restored.

The stronger and cleaner play is to anchor your DMCA claim on whatever human authorship exists. If you wrote the lyrics yourself, those lyrics are your copyrighted literary work regardless of who sang them or what generated the backing track, and the stolen upload reproduces them without permission. That is a legitimate, defensible takedown. Same if you edited the AI output in a DAW, a digital audio workstation like Ableton or Logic, by rearranging sections, layering your own vocals, or adding recorded instruments. Forensic musicologist Brian McBrearty of Musicologize frames it simply: the portions you created yourself remain protectable, so if the track was half you and half AI, your half is still yours.

Practically, here’s the sequence. First, screenshot everything about the infringing release: artist name, track title, ISRC code if visible, upload date, stream counts. Second, find the right door. For Spotify, use their online infringement form. For YouTube, the copyright complaint flow. If the stolen track went out through a distributor you can identify, send the notice to the distributor too, since they can pull the release from every platform at once. Third, send the notice identifying your original work, the infringing material, and your good-faith statement. If the uploader files a counter-notice, the platform typically restores the content after 10 to 14 business days unless you file suit, so mark that window on your calendar and decide in advance how far you’ll go.

One more tool that hardly anyone uses: the DMCA subpoena. Section 512 of the Act lets a copyright owner ask a federal district court clerk to order a platform or distributor to hand over the uploader’s identifying information. Lyman-Schmidt used exactly this provision in his fight, as he described in the Philadelphia Magazine piece. It requires a real copyright claim, so again, it works best when human-authored elements exist. But if someone built a business on stealing your catalog, knowing who they are changes the conversation entirely.

The Platform Policies That Changed the Game

If the DMCA is a scalpel, the 2025 wave of platform policy changes is a sledgehammer, and I’d argue it’s now the more reliable route for AI creators specifically, because it doesn’t hinge on copyright at all.

Spotify announced on September 25, 2025 that it had removed more than 75 million spammy tracks over the preceding 12 months and rolled out three new pillars: an impersonation policy with clearer recourse for artists, a music spam filter targeting mass uploads and duplicates, and support for standardized AI disclosures in credits through DDEX, the industry metadata standard. For your situation, the impersonation and content mismatch provisions are the gold. Content mismatch is Spotify’s term for fraudulent uploads delivered to the wrong artist profile, and the company said it has partnered with distributors to stop these attacks at the source, cut review wait times, and let artists report mismatches even before a release goes live. Consequence and Forbes both covered the details, and Spotify’s enforcement has had visible teeth since; when AI-generated fakes appeared on Anne Murray’s profile in January 2026, Billboard Canada reported they were pulled within four days.

What this means for you: if the thief uploaded your track to your artist profile, or under a name confusingly similar to yours, report it as impersonation or mismatch through Spotify for Artists rather than, or in addition to, a DMCA notice. These reports are evaluated against platform rules, not against the question of whether AI audio can be copyrighted, and that distinction works in your favor. The same logic applies at Apple Music, YouTube Music, and Deezer, which built its own AI-detection tagging system.

Your distributor is the other pressure point. If you had already scheduled your own release through DistroKid, TuneCore, CD Baby, or similar, tell them immediately that a duplicate of your master is live from another source, and give them your generation timestamps. Distributors talk to platforms through channels you can’t access, and duplicate-master disputes are routine business for them. Fair warning from the trenches, though: response quality varies wildly. In one Spotify Community thread on false claims, artists reported TuneCore reinstating music promptly after a counter-notice while Symphonic left tracks down even after the claimant went silent. Choose a distributor with a human support team before you need one.

The Human Element Strategy for Everything You Make Next

Everything above is damage control. The durable fix is to stop releasing songs that have no copyright attached to any part of them.

The Copyright Office’s January 2025 report didn’t just close doors; it opened one, clearly. Human contributions that are perceptible in AI-assisted works are protectable, and so is your creative selection, coordination, arrangement, and modification of AI material. The Office noted it had already registered more than a thousand works where applicants disclosed and disclaimed the AI-generated portions, per analysis from Rimon Law. Generate 50 melody fragments, then select, arrange, and rework three of them into a finished song, and you’ve likely crossed the authorship threshold, as Orphiq’s guide illustrates. Generate ten finished tracks and just pick your favorite, and you likely haven’t.

My personal workflow, for what it’s worth. I write every lyric myself in a text file before I ever open Suno, because lyrics are the cheapest, strongest copyright you can attach to an AI song. I pull stems and rearrange the structure in my DAW, mute and replace at least one section with something I played or sang, and export the final master myself. When I register with the Copyright Office, I disclose the AI-generated material and claim the human parts, exactly as their guidance requires. Registration costs $45 to $65 per work through eCO, the electronic Copyright Office system, and for anything I care about commercially, it’s the best money in my budget. Registration before infringement is what unlocks statutory damages and attorney’s fees in a US lawsuit, which is the difference between a threat a thief can ignore and one they can’t.

And declare your AI use honestly when you distribute. Every major platform now supports DDEX AI disclosure fields through distributors, and undisclosed AI is increasingly what gets releases pulled and accounts reviewed, a point the composer behind Silverman Sound makes forcefully in his legal-risk guide for creators. Hiding the AI doesn’t protect you from thieves. It just adds a second way to lose your catalog.

Build Your Evidence Vault Before You Need It

When my track got jacked, the thing that saved me wasn’t a law. It was a timestamp. My Suno generation history showed the creation date, my Discord post predated the thief’s release, and my lyrics file had a file-creation date from a week earlier. That stack of boring evidence is what convinced the distributor to act.

Do this for every track from now on. Keep the prompt, the generation date, and a screenshot of your platform library entry. Save your lyric drafts and voice memos with their original file dates. Email the finished file to yourself or store it in cloud storage that logs upload times, since those third-party timestamps are harder to dispute than anything on your own hard drive. Some creators go further with cryptographic registration services that hash the audio file and issue a signed, timestamped certificate; SoundGuard AI, one of the newer entrants, markets exactly this to Suno and Udio users. I’m mildly skeptical that any private registry certificate carries weight in court on its own, and none of them substitutes for actual Copyright Office registration of your human elements. But as persuasive evidence in a distributor or platform dispute, a hash plus a timestamp is genuinely useful, and it costs you five minutes.

One more habit that hurts to recommend, because sharing is half the fun: stop posting full-quality, unreleased tracks in public spaces. Post snippets. Post the track with a spoken tag over it. Or distribute first and share second, so that every automated system already associates the master with your name. The window between when you share and when you release is the entire attack surface. Shrink it.

What I’d Do in the First 48 Hours

If the theft just happened, here’s the exact order I’d work in, based on going through it myself.

Hour one, preserve evidence. Screenshot the infringing release everywhere it appears, capture your own generation history and timestamps, and export your Suno or Udio library entry before anything changes. If you’re on a free tier, confirm right now whether you even hold commercial rights to the track, because that shapes every claim you make afterward.

Day one, hit the platforms. File a content mismatch or impersonation report through Spotify for Artists if the release touches your profile or name. File infringement reports on each platform where the track appears, anchored on your human-authored elements if you have them. Be factual and unemotional in every form field; the person reading it processes hundreds of these.

Day two, chase the source. Identify the thief’s distributor from the release metadata or the label line on streaming pages, and send your notice and evidence there. Notify your own distributor if you have one. If real money is flowing to the stolen version, this is also the moment to spend a few hundred dollars on a consult with a music attorney about a Section 512 subpoena and whether your human elements support registration and a stronger claim. Nothing in this article is legal advice, and the AI music copyright landscape shifts monthly; a professional reading your specific facts is worth it once dollars are at stake.

Then document everything and follow up weekly. Platforms lose reports. Distributors go quiet. The squeaky wheel genuinely gets the takedown.

Quick Answers to the Questions You’re Probably Asking

Can the thief sue me if I release my own version after theirs? Almost certainly not on copyright grounds, for the same reason you can’t sue them. If the audio is fully AI-generated, they hold no copyright in it either, and being first to upload doesn’t create one. Their distributor’s duplicate detection might flag your release, which is annoying, but that’s a dispute you win with your generation timestamps, not a lawsuit you lose.

Should I still release my version? Usually yes, and quickly, through a distributor that lets you attach your evidence if a conflict arises. Holding your release hostage to their takedown timeline just hands them more streams. The exception is if their version has gone genuinely viral; then talk to a lawyer before doing anything, because there may be revenue worth fighting over properly.

Does any of this change outside the US? Somewhat. The Copyright Office noted an emerging international consensus around human authorship, and most major markets currently point the same direction, though a July 2026 German court ruling against Suno over training data shows how fast the platform side of the law is moving in Europe. The takedown mechanics in this article are US-centric, but Spotify’s impersonation and mismatch policies apply globally, which is one more reason to lean on them first.

An Honest Reality Check

I’ll close with the part I wish someone had told me at the start. Sometimes you do everything right and the stolen version stays up anyway, or gets restored after a counter-notice you can’t afford to fight in federal court. If your song was purely AI-generated, with no lyrics or edits of yours in it, the law as it stands in 2026 simply does not give you ownership of that audio, and no amount of righteous anger changes the Copyright Office’s position. The Warner and Universal settlements are pushing Suno and Udio toward licensed, walled-garden models that may make ripping harder, but they’ve also made user rights thinner, not thicker.

The creators I see thriving in AI music treat the generator as one instrument in a human process, keep meticulous records, register what’s registrable, and release fast. The ones I see getting burned repeatedly treat the AI output as a finished product and the internet as a friendly place. If someone released your AI song before you, you’ve already paid the tuition. This week, pick your best unreleased track, write or rewrite the lyrics yourself, timestamp everything, and get it distributed under your own name before anyone else can. That’s the version of ownership that’s actually available right now, and it’s more than enough to build on.

Sources

Common questions

Can I sue someone who released my AI-generated song before me?

Probably not on copyright grounds for the audio, because a song generated entirely by AI from a prompt has no US copyright protection since copyright requires a human author. You still have other options, including takedowns, platform impersonation policies, and copyright claims on any human elements you contributed.

Why do AI songs get stolen so often?

When large volumes of synthetic audio flood distribution pipelines, opportunists follow, and raw AI files carry no embedded project data, stems, or session files tying them to you. Theft usually happens at the sharing stage, when someone rips a track you posted for feedback and uploads it to a cheap distributor within days, sometimes speeding it up or down to dodge duplicate detection.

Do I own the song if I made it on Suno or Udio?

Your rights come from the platform's contract, not copyright, and only paid-tier songs carry commercial rights, while free-tier songs are for personal use only and are not retroactively upgraded by subscribing later. After the Warner and Universal settlements the platforms shifted their language from ownership to granted commercial rights and warn that no copyright is guaranteed to vest in your output.

How do I file a DMCA takedown for a stolen AI track?

Screenshot everything about the infringing release, find the right reporting form for the platform or distributor, and send a notice identifying your original work with a good-faith statement. The claim is strongest when anchored on human-authored elements like lyrics you wrote or edits you made, since a pure claim on fully AI-generated audio stands on shaky ground.

Which streaming platform policies help with a stolen AI song?

Spotify's September 2025 changes added an impersonation policy and a content mismatch process for fraudulent uploads delivered to the wrong artist profile, which are evaluated against platform rules rather than whether AI audio can be copyrighted. If the thief used your profile or a confusingly similar name, report it as impersonation or mismatch through Spotify for Artists, and similar logic applies at Apple Music, YouTube Music, and Deezer.

How can I make my AI songs copyrightable going forward?

Add perceptible human contributions, since the Copyright Office says your creative selection, arrangement, and modification of AI material is protectable, and writing your own lyrics is the cheapest and strongest copyright you can attach. Generating many fragments and then selecting, arranging, and reworking a few likely crosses the authorship threshold, whereas just picking a favorite from finished tracks likely does not.

What evidence should I keep to protect my AI tracks?

Keep the prompt, generation date, and a screenshot of your platform library entry, and save lyric drafts and voice memos with their original file dates. Emailing the finished file to yourself or using cloud storage that logs upload times creates third-party timestamps that are harder to dispute, and these boring records are often what convinces a distributor to act.

Should I still release my own version after someone stole it?

Usually yes, and quickly, through a distributor that lets you attach your evidence if a conflict arises, since holding your release back just hands the thief more streams. The exception is if their version has gone genuinely viral, in which case you should talk to a lawyer first because there may be real revenue worth fighting over.

How can I stop my AI songs from being stolen in the first place?

Stop posting full-quality unreleased tracks in public, and instead post snippets, add a spoken tag over the audio, or distribute first and share second so automated systems already associate the master with your name. The window between sharing and releasing is the entire attack surface, so shrinking it is the most effective prevention.