Plain-language resources for labels, publishers, distributors and catalog owners - on how platforms actually work, what the law requires, and how to protect a catalog in the era of AI-generated music.
Why YouTube and Meta allow claiming and monetisation while streaming services only allow removal - and how to set a catalog policy that uses each channel correctly.
What "exclusive rights to a substantial body of original material" means in practice, why access is restricted, and how rights-management partners operate inside it.
Audio and video matching across Facebook and Instagram, the block / track / monetise actions, the approval process, and the consequences of misuse.
The six required elements of a takedown notice, the counter-notification process, the 10-14 business day window, and why §512(f) matters to anyone sending notices at scale.
Best-efforts obligations, stay-down, Article 16 notices and trusted-flagger status - and what they mean when you enforce in Europe.
Why a cover in a different key with a different singer is still your song, how composition-level matching differs from recording-level matching, and what you can do about it.
A sober reading of published platform data on AI-generated uploads and fraudulent streams, with sources - and what it means for monitoring strategy.
The metadata and documentation that makes enforcement fast and accurate - and the gaps that cause false claims.
How to read a dispute, when to release, when to hold, and why escalation should be rare and deliberate.
Master vs. composition, ISRC vs. ISWC, Content ID vs. Rights Manager, claim vs. takedown vs. counter-notice, chain of title, stay-down, and more - defined plainly.
Nothing on this page is legal advice; for decisions about your catalog, speak to counsel.
We do not publish a single headline accuracy number. We publish how we measure: per-category performance (exact re-uploads, edits, covers, AI derivatives), how reviewers confirm matches, and how disputes feed back into our metrics. The methodology page is updated with every model release.
When a platform changes what rights holders can do, we write it up. No marketing, no more than once a month.
Why YouTube and Meta allow claiming and monetisation while streaming services only allow removal - and how to set a catalog policy that uses each channel correctly.
Rights holders often arrive with one question: "Can you take it down?" The better question is: "What does this platform actually let a rights holder do?" The answer differs sharply by platform type, and a catalog policy that ignores those differences either leaves money on the table or sends notices that cannot succeed.
1. Rights-management programs. A small number of platforms run programs that let approved rights holders deliver reference files, match uploads automatically, and apply a policy to each match: monetise (run ads and collect revenue), track (collect analytics), or block. YouTube's Content ID and Meta's Rights Manager are the two that matter most. Access is granted by application and can be revoked for misuse.
2. Statutory notice-and-takedown. Most platforms - streaming services, download stores, social networks without a rights program, live platforms, hosts - provide a process for sending a legally compliant notice identifying infringing material. The platform then removes or disables the material. There is no monetisation; the remedy is removal (or, in some cases, territory restriction).
3. Distributor-level withdrawal. On streaming services and download stores, content arrives via a distributor. If the infringing release was delivered by your distributor, or by a distributor you can reach, withdrawal through that channel is usually faster than a public infringement form.
| Platform type | Typical channel | Available outcomes |
|---|---|---|
| YouTube & Shorts | Rights-management program (Content ID), subject to eligibility | Monetise · Track · Block |
| Facebook & Instagram | Rights-management program (Rights Manager), subject to approval | Monetise (ad earnings) · Track · Block |
| TikTok, Snapchat, X, Reddit, Pinterest | IP reporting process | Removal |
| Spotify, Apple Music, Amazon Music, Deezer, Tidal | Infringement forms; distributor withdrawal | Removal · territory disable |
| iTunes Store, Beatport, Bandcamp, Qobuz | Store forms; distributor withdrawal; label contact | Removal |
| SoundCloud, Twitch | Reference-database registration; copyright forms; DMCA | Block / mute at upload · Removal |
| Cyberlockers, torrent indexes, stream-rippers | DMCA / DSA notices; search delisting | Takedown · Delisting |
A good policy is written per platform and per use type, not as one global rule:
No third party - Sonproof included - can monetise a use on a platform that has no rights program, or guarantee removal on a platform that evaluates notices itself. Anyone promising "instant removal everywhere" is describing something that does not exist. What a good partner can do is use every channel a platform provides, correctly, every time, and tell you honestly which one applies. See our Copyright Enforcement Policy for how we do that.
What "exclusive rights to a substantial body of original material" means in practice, why access is restricted, and how rights-management partners operate inside it.
Content ID is YouTube's reference-matching system. Rights holders deliver audio or video references; uploads are scanned; matches receive the rights holder's policy - monetise, track or block. It is the single largest mechanism by which rights holders earn from user uploads: YouTube has reported cumulative Content ID payouts in the billions of dollars, and processes billions of claims a year (Google Transparency Report).
YouTube's published criteria say Content ID is for rights holders who own exclusive rights to a substantial body of original material that is frequently uploaded by users, and who have the resources to manage it responsibly. In practice that excludes most individual creators and many small labels, and it excludes content for which the applicant holds only non-exclusive rights (a common situation for sample packs, libraries and some distribution deals). Applicants sign an agreement restricting references to content they exclusively own and declaring the territories covered.
Because the system acts at scale and automatically, a bad reference - a file the claimant does not exclusively own, a public-domain work, a licensed sample - generates invalid claims against thousands of legitimate videos. YouTube therefore gates access and revokes it for repeated invalid claims. Misuse damages not just the offending partner but the credibility of the whole system.
Many rights holders access Content ID through a partner that already holds a Content Manager account. The partner delivers references on the rights holder's behalf, applies the rights holder's policy, handles disputes, and reports revenue. A responsible partner reviews references before delivery, maintains whitelists, and releases claims promptly when a dispute has merit - because the partner's own access depends on accuracy across all of its clients.
Uploaders can dispute a claim. The claimant has a review window (YouTube's help pages describe a 30-day period) to release or uphold. Upholding a disputed claim, and especially escalating to a formal takedown, should be rare and deliberate; it is where reputational and legal risk concentrates. See our Counter-Notice Policy.
Sources: YouTube Help - "Qualifying for Content ID"; Google Transparency Report - YouTube copyright.
Audio and video matching across Facebook and Instagram, the block / track / monetise actions, the approval process, and the consequences of misuse.
Rights Manager is Meta's tool for rights holders to find matches to their content across Facebook and Instagram and apply actions to them. It covers video, audio and image matching, and is the channel through which labels and publishers can claim and, where eligible, monetise uses of their music on those platforms.
Access is by application through Meta's business tools and is discretionary. Meta expects applicants to own or control exclusive rights to a catalog of original content and to use the tool responsibly. Meta has publicly tightened scrutiny after reports in 2024 of Rights Manager being abused to extort creators; it states that misuse can result in loss of access to Rights Manager and other Meta products.
Deliver only references you exclusively control; keep your whitelist (your own pages, partners, licensed uses) current; review matches before applying block actions at scale; and handle disputes on their merits. As with Content ID, the tool rewards accuracy and penalises carelessness.
Sources: Meta Business Help Center - Rights Manager overview and eligibility; Rights Manager Terms.
The six required elements of a takedown notice, the counter-notification process, the 10-14 business day window, and why §512(f) matters to anyone sending notices at scale.
This is an explainer, not legal advice. For decisions about your catalog, speak to counsel.
Section 512 of the US Copyright Act (17 U.S.C. §512) creates "safe harbors" that shield online service providers from monetary liability for user infringement, provided they meet certain conditions - including responding to compliant takedown notices. It is the legal backbone of notice-and-takedown in the United States and shapes how most platforms worldwide handle notices.
A notice missing these elements may be treated as non-compliant, and the provider may have no obligation to act on it.
A user whose content was removed may send a counter-notification to the provider's designated agent containing: their signature; identification of the removed material and its former location; a statement under penalty of perjury of good-faith belief that removal was a mistake or misidentification; and their name, address, phone number, consent to federal court jurisdiction, and acceptance of service of process. The provider forwards it to the original complainant and, unless the complainant notifies the provider within 10 to 14 business days that it has filed a court action, the provider restores the material.
Providers seeking the safe harbor must designate an agent to receive notices, both on their website and in the US Copyright Office's DMCA Designated Agent Directory. Notices should be sent to that agent, not to generic support addresses.
Anyone who knowingly materially misrepresents that material is infringing (in a notice) or was removed by mistake (in a counter-notice) is liable for damages, including costs and attorney's fees, incurred by the alleged infringer, the copyright owner, or the service provider. For anyone sending notices at scale, this is the core legal risk - and the reason every notice should be verified by a human before it is sent.
How we apply this in practice is described in our Copyright Enforcement Policy.
Sources: 17 U.S.C. §512; U.S. Copyright Office, Section 512 resources and DMCA Designated Agent Directory.
Best-efforts obligations, stay-down, Article 16 notices and trusted-flagger status - and what they mean when you enforce in Europe.
This is an explainer, not legal advice.
Two EU instruments reshaped enforcement against online platforms in Europe: Article 17 of the Copyright Directive (Directive (EU) 2019/790) and the Digital Services Act (Regulation (EU) 2022/2065). They work differently and apply to different kinds of platform, and rights holders enforcing in the EU should understand both.
Article 17 applies to "online content-sharing service providers" (OCSSPs) - broadly, large platforms whose main purpose is to host and give public access to large amounts of user-uploaded copyright content (YouTube, Meta's platforms, TikTok, SoundCloud and similar are in scope; pure streaming stores like Spotify or Apple Music, where content is delivered by labels and distributors rather than users, are generally not).
Under Article 17 an OCSSP performs an act of communication to the public when it gives access to user uploads, and is liable for unauthorised uploads unless it shows that it (a) made best efforts to obtain authorisation (licences); (b) made best efforts, in accordance with high industry standards, to ensure the unavailability of specific works for which rights holders have provided the relevant and necessary information (in practice, reference files and rights data); and (c) acted expeditiously on notice to remove content and made best efforts to prevent its future upload (stay-down).
For rights holders, the practical effect is that supplying good reference data to OCSSPs gives you proactive blocking and stay-down, not only reactive takedowns - which is why program access (Content ID, Rights Manager and equivalents) is so valuable in the EU. The Court of Justice (Case C-401/19, 2022) upheld Article 17 while stressing that filtering must not block lawful uploads, and the Commission's guidance (COM/2021/288) emphasises safeguards for exceptions such as quotation, criticism and parody.
The DSA applies to all intermediaries, including hosting providers that are not OCSSPs. Its Article 16 requires hosting providers to put in place easy-to-use, electronic notice-and-action mechanisms through which any person or entity can flag specific items of allegedly illegal content - including copyright infringement. A notice should explain why the content is illegal, give its exact location (URL), identify the notifier, and include a good-faith statement of accuracy. A sufficiently precise and substantiated notice can give the provider actual knowledge for liability purposes.
The DSA also requires providers to give a statement of reasons when they restrict content (Art. 17), to operate an internal complaint-handling system (Art. 20), and to allow referral to certified out-of-court dispute settlement bodies (Art. 21) - the EU analogue of counter-notice. And it creates trusted flaggers (Art. 22): entities awarded that status by a national Digital Services Coordinator whose notices platforms must process with priority.
Sources: Directive (EU) 2019/790, Art. 17; European Commission Guidance COM/2021/288; CJEU Case C-401/19; Regulation (EU) 2022/2065, Arts. 16, 17, 20, 21, 22.
Why a cover in a different key with a different singer is still your song, how composition-level matching differs from recording-level matching, and what you can do about it.
Every recorded track carries two separate copyrights. The sound recording (the "master") is the specific performance captured in the file - owned typically by a label or the artist. The musical composition (the "work") is the underlying melody, harmony and lyrics - owned typically by songwriters and their publisher. A cover version is a new sound recording of the same composition; it does not use the master, but it does use the work.
Classic audio fingerprinting identifies recordings. It is extremely robust to re-encoding, noise, speed and pitch shifts, and edits - but it is built to recognise the same performance. A cover by a different singer in a different key at a different tempo is, acoustically, a different recording, and a recording-level fingerprint will usually not match it. That is by design, and it is why publishers have historically had far less visibility than labels.
Composition-level (or "version") identification models the features that survive re-performance: melodic contour, harmonic progression, rhythmic structure, and - where available - lyrics. Rather than asking "is this the same audio?", it asks "is this the same song?" It is harder, it produces more candidates that need review, and its confidence is expressed differently (a cover can be a strong match on melody and weak on everything else). Several vendors and platforms have shipped version-identification systems in recent years; the research field calls the problem "cover song identification".
AI-generated music complicates both layers. An AI vocal swap keeps the recording's instrumental and replaces the voice; a "remake" regenerates the whole track from a prompt; a derivative borrows melody or structure. Some of these match at recording level, some only at composition level, some are best identified by separate AI-detection classifiers. Public platform data shows the scale: Deezer has reported that fully AI-generated tracks exceeded half of its daily uploads by mid-2026, with the majority of their streams flagged as fraudulent (Deezer Newsroom, 2026). Detecting a derivative of your work inside that flood is a composition-level problem.
A match score does not establish infringement. It establishes similarity. Whether a similar work infringes depends on originality, substantial similarity, licences, and exceptions - questions for rights teams and, sometimes, courts. A responsible system keeps those two things apart: detection surfaces, people decide.
Sources: Deezer Newsroom (2026) on AI-generated uploads; research literature on cover song identification; platform program documentation.
A sober reading of published platform data on AI-generated uploads and fraudulent streams, with sources - and what it means for monitoring strategy.
It is easy to talk about AI music in superlatives. This piece sticks to figures that platforms and reputable outlets have published, and says what they do and do not show.
Three things. First, streaming services now matter as enforcement targets, not only UGC platforms: impersonation, unauthorised re-releases and sped-up/AI re-uploads arrive through distribution and need distributor-level and form-based action. Second, composition-level detection is what finds derivatives of your works inside the flood. Third, the regulatory direction - EU AI Act transparency rules applying from August 2026, and US proposals such as the NO FAKES Act - points toward more labelling and more accountability, which helps rights holders who can document what is theirs.
Sources: Deezer Newsroom (Apr and Jul 2026); Spotify newsroom (Sep 2025) and Music Business Worldwide coverage; U.S. Department of Justice, SDNY press release (Sep 2024) and 2026 plea coverage; European Commission, AI Act Article 50 guidance (2026).
The metadata and documentation that makes enforcement fast and accurate - and the gaps that cause false claims.
Most enforcement problems are not detection problems. They are data problems: a reference delivered for a recording the client does not exclusively own, a territory that was carved out in a distribution deal, a licensed partner missing from a whitelist. Fixing the data first makes everything downstream faster and safer.
The single most common cause of false claims is a licensed use that nobody told the enforcement system about. Build and maintain a whitelist: your own channels and accounts; licensed partners and their channels; sync licensees; promotional grants; settlements. Update it when deals change.
Deliver clean, full-length masters (not radio edits only, not low-bitrate previews) so references cover the whole recording. For compositions, a reference recording per work plus lyrics where available improves version matching.
For large catalogs, DDEX-based delivery or a structured CSV export from your catalog system is faster and less error-prone than manual upload. Expect the enforcement partner to validate the data and come back with questions before anything is scanned - that is a good sign.
Ready to onboard? Book a demo and bring a few ISRCs.
How to read a dispute, when to release, when to hold, and why escalation should be rare and deliberate.
Disputes are not attacks. Most are one of a handful of things, and each has a right response. A rights team that treats every dispute as a fight will eventually send the notice that gets it sued; a team that treats every dispute as a question will rarely be wrong.
Before anything else, establish which of these it is:
Whenever any of 1-3 is true, or when 4 is plausible and you cannot confidently rebut it. Releasing a weak claim costs you nothing; holding it can cost you a platform relationship, a §512(f) claim, or a public incident.
Only when the match is genuine, your rights clearly cover the use, no licence or exception applies, and the record shows you checked. Write down why.
Converting a program dispute into a formal takedown, or responding to a DMCA counter-notice by notifying the platform of a court action, are serious steps. They should require a second reviewer, client instruction, and a genuine intention to pursue the matter. Escalation as a negotiating tactic is how rights holders end up on the wrong side of misrepresentation claims.
Record the outcome and the reason in the case file; feed false positives back into matching; feed missed licences back into the whitelist. A dispute you learned from is a dispute that paid for itself.
How Sonproof handles disputes is set out in our Counter-Notice Policy.
Master vs. composition, ISRC vs. ISWC, Content ID vs. Rights Manager, claim vs. takedown vs. counter-notice, chain of title, stay-down, and more - defined plainly.