The DMCA takedown notice: how it works and what it cannot do

The Digital Millennium Copyright Act added section 512 to Title 17 of the United States Code, and with it the notice-and-takedown procedure that now governs a very large share of content removal worldwide. It is worth understanding precisely, because it is both more useful and more limited than its reputation suggests.

What section 512 actually is

Section 512 does not create liability. It creates limitations on liability — safe harbours — for service providers who meet its conditions. There are four:

Only (c) and (d) carry the notice-and-takedown machinery. Subsection (a) has no takedown mechanism at all — a conduit stores nothing and has nothing to remove. This is the most common structural misunderstanding in practice, and it produces a great many notices sent to the wrong recipient.

Note also that even a provider which retains its safe harbour remains subject to the limited injunctive relief available under section 512(j). Safe harbour is not immunity from every order.

Separately, section 512(i) conditions safe harbour under any subsection on the provider having adopted and reasonably implemented a policy for terminating repeat infringers in appropriate circumstances, and on accommodating standard technical measures.

Territorial reach — and why a Kenyan claimant can use it

Section 512 is a limitation on liability under United States copyright law. It applies because the service provider is exposed to US copyright claims — that is, because of where the platform, host or its operations sit — not because of the claimant’s nationality, residence, or where the harm was felt.

Three consequences follow:

• A Kenyan or Emirati rightsholder can send a valid notice to a US-based host. Under the national-treatment principle in Article 5(2) of the Berne Convention their work is protected in the United States automatically, with no formality required to send a notice.
• US registration is not a precondition to a notice. It is required to file suit in respect of a US work, but that is a different step.
• A notice sent to a non-US host has no statutory force. Many honour “DMCA notices” voluntarily or by contract; that is grace, not law. Against an EU host the correct instrument is a notice under Article 16 of the Digital Services Act; against a Kenyan ISP, the section 35B regime described in Who owns online content.

The six elements of a valid notice

Section 512(c)(3)(A) requires a written communication to the provider’s designated agent, including substantially the following:

1. A physical or electronic signature of a person authorised to act for the owner of an exclusive right allegedly infringed.
2. Identification of the copyrighted work claimed to have been infringed, or a representative list where multiple works at a single site are covered.
3. Identification of the infringing material, with information reasonably sufficient to permit the provider to locate it.
4. Information reasonably sufficient to permit the provider to contact the complaining party.
5. A statement that the complaining party has a good faith belief that the use is not authorised by the copyright owner, its agent, or the law.
6. A statement that the information is accurate, and — under penalty of perjury — that the complaining party is authorised to act on behalf of the owner.

A drafting point that a great deal of loose commentary reverses: the penalty-of-perjury oath attaches to element 6 — accuracy and authority — not to the good-faith belief in element 5.

Under section 512(c)(3)(B), a notice that fails to comply substantially does not confer actual knowledge on the provider. But where a notice substantially complies with elements 2, 3 and 4 only, the provider must promptly attempt to contact the sender or take other reasonable steps to help obtain a compliant notice before disregarding it.

Check the designated agent first. Designations must be made through the Copyright Office’s online directory; paper designations expired on 1 January 2018. A designation costs six dollars, is valid for three years, and must be renewed. A host with a lapsed or missing designation cannot claim the section 512(c) safe harbour at all, which changes the negotiating position entirely.

Counter-notification and the put-back window

This is the part of the process most likely to surprise a claimant.

A provider that removes material in good faith is protected under section 512(g)(1) “regardless of whether the material or activity is ultimately determined to be infringing”. The statute rewards removing first and asking later, which is why platforms do exactly that.

The subscriber may then file a counter-notice under section 512(g)(3), which requires four things: a signature; identification of the material and where it appeared; a statement under penalty of perjury of a good-faith belief that the material was removed as a result of mistake or misidentification; and the subscriber’s contact details together with a statement consenting to the jurisdiction of a United States Federal District Court — for the district of their address, or, if the subscriber’s address is outside the United States, for any judicial district in which the service provider may be found — and agreeing to accept service of process.

That jurisdictional consent is significant and is routinely overlooked. Filing a counter-notice from Nairobi or Dubai submits the sender to US federal jurisdiction. Anyone considering one should understand that before signing it.

If a counter-notice is filed, the provider must restore the material not less than 10 and not more than 14 business days after receipt, unless its designated agent first receives notice that the original complainant has filed an action seeking a court order restraining the infringing activity. Filing suit stops the clock. Threatening to does not.

The practical consequence: a claimant who sends a notice they are not prepared to litigate has, at best, bought two to three weeks — and has put on record a copyright claim they were unwilling to stand behind.

Misrepresentation liability

Section 512(f) makes any person who knowingly materially misrepresents that material is infringing, or that it was removed by mistake, liable for damages, costs and attorneys’ fees incurred by the alleged infringer, the copyright owner or licensee, or the service provider.

Two authorities frame the risk. In Online Policy Group v. Diebold, Inc., 337 F. Supp. 2d 1195 (N.D. Cal. 2004), the court held that no reasonable copyright holder could have believed the material at issue — an email archive discussing problems with the defendant’s voting machines — was protected, and the company paid $125,000 in damages and fees. In Lenz v. Universal Music Corp., 801 F.3d 1126 (9th Cir. 2015), amended at 815 F.3d 1145 (9th Cir. 2016), the court held that a rightsholder must consider fair use before sending a notice, though it applied a subjective standard under which a genuine, even unreasonable, belief escapes liability — with wilful blindness capable of substituting for knowledge.

Section 512(f) is a real but narrow hook. It bites hardest where the claimant plainly has no copyright interest in the material at all — which is precisely the shape of a reputational grievance dressed as a copyright claim.

The identification subpoena, and its limit

Section 512(h) allows a copyright owner or its agent to ask the clerk of any US district court to issue a subpoena requiring a service provider to identify an alleged infringer. It requires a compliant notification, a proposed subpoena, and a sworn declaration that the purpose is identification and that the information will be used only to protect rights under the Copyright Act. The clerk issues it without a judge or an adversarial hearing.

It is an attractive route, and it does not work against a mere conduit. Three federal circuits now agree that a section 512(h) subpoena cannot issue to an access ISP acting solely under section 512(a): RIAA v. Verizon Internet Services, 351 F.3d 1229 (D.C. Cir. 2003); In re Charter Communications, 393 F.3d 771 (8th Cir. 2005); and In re Subpoena of Internet Subscribers of Cox Communications, No. 24-3978 (9th Cir., 15 August 2025), which held as a matter of law that the subpoena presupposes material the recipient can remove or disable, and a pure conduit has none.

Where the only intermediary is an access provider, identification therefore requires a different route — a John Doe action with subpoenas under the ordinary rules, or, for foreign proceedings, an application under 28 U.S.C. section 1782.

What a notice cannot do

• It cannot resolve a defamation claim. Copyright protects the expression of a work; it says nothing about whether the statements in it are true, fair or damaging.
• It cannot remove a photograph someone else took of you. The photographer is the author. Being in the picture confers no copyright.
• It cannot remove an article written about you. The author owns it.
• It cannot be undone quietly. Notices are routinely published — the Lumen Database receives them from major platforms, and Google’s transparency reporting publishes URL-level data. A notice intended to bury a story can create a permanent, indexed, searchable record of the attempt, under the claimant’s own name.

Where copyright genuinely applies — a photograph the client took, their own writing, artwork, video or music reposted without licence, or commissioned work with a documented assignment — it is an efficient and effective mechanism. Where it does not, there are better instruments, and using the wrong one has a cost.