Mavrix v. LiveJournal: When Volunteer Moderators Put the DMCA Safe Harbor at Risk
The Ninth Circuit held volunteer moderators may be a platform's agents, so posts they screen may fall outside DMCA 512(c)'s user-direction safe harbor.
The DMCA safe harbor is 17 U.S.C. § 512, a set of four liability shields that protect qualifying online service providers from monetary damages for copyright infringement committed by their users. It is the legal architecture that makes user-generated content possible. Without it, every host that ever stored an uploaded file would face statutory damages of up to $150,000 per work for material it never chose, saw, or wanted.
Two things about the shield are widely misunderstood. First, it is not a finding of innocence. A provider inside the safe harbor may have facilitated real infringement; the statute simply directs the copyright owner to sue the uploader rather than the platform. Second, falling out of the safe harbor does not make a provider liable. It only means the provider must defend on the ordinary law of secondary liability, where it may still win.
Congress did not write one shield. It wrote four, matched to four functions a service can perform, and a provider gets the protection for the function it is actually performing.
The distinction is not academic. A cable ISP served with a takedown notice cannot remove the file, because it never had the file. Copyright owners have repeatedly tried to treat 512(a) conduits as though they owed 512(c) duties, and the statute does not support it.
Before any of the four shields is available, § 512(i) imposes two conditions on the provider as a whole. It must adopt, reasonably implement, and inform subscribers of a policy of terminating repeat infringers in appropriate circumstances. It must also accommodate standard technical measures used to identify or protect works.
For the hosting and linking harbors, § 512(c)(2) adds a registration step that is easy to satisfy and easy to fumble: designate an agent to receive notices, publish that agent’s contact information on the site, and file the designation with the Copyright Office. Since December 2016 that filing is electronic, costs $6, and expires after three years unless renewed. A lapsed registration is a paperwork failure with catastrophic downside, because the shield is unavailable during any period the designation is not on file.
The hosting harbor evaporates if the provider has actual knowledge of specific infringing material, or is aware of “facts or circumstances from which infringing activity is apparent,” the provision courts call red flag knowledge. The fight has always been over how specific that knowledge must be.
Viacom International, Inc. v. YouTube, Inc., 676 F.3d 19 (2d Cir. 2012), settled the framework. The Second Circuit held that both actual and red flag knowledge must be item-specific. General awareness that infringement is rampant on your service, even internal emails estimating that most of the catalog is unlicensed, does not disqualify you. The knowledge must attach to particular material. The Ninth Circuit reached the same conclusion in UMG Recordings, Inc. v. Shelter Capital Partners LLC, 718 F.3d 1006 (9th Cir. 2013), and the Second Circuit reinforced it in Capitol Records, LLC v. Vimeo, LLC, 826 F.3d 78 (2d Cir. 2016), holding that a staffer watching part of a video containing recognizable music does not, without more, make infringement objectively obvious.
Two limits cut the other way. Willful blindness counts: a provider that deliberately looks away from specific infringement it suspects can be charged with knowledge. And § 512(m) confirms the flip side, that the shield does not condition itself on affirmatively monitoring your service for infringement. You need not go looking. You may not stage-manage not knowing.
Section 512(c)(1)(B) removes the shield from a provider that receives “a financial benefit directly attributable to the infringing activity” while having “the right and ability to control” it. Read literally, every ad-supported host with a delete button would fail this test, which would swallow the statute whole.
Courts refused that reading. Viacom and UMG both held that control under § 512(c) means something more than the ordinary ability to remove content or enforce terms of service. It requires substantial influence over the infringing activity, of the kind seen where an operator induces infringement or actively curates and directs what gets uploaded. Ordinary moderation is not control.
If the safe harbor fails, it usually fails here, on the least glamorous condition in the statute.
BMG Rights Management (US) LLC v. Cox Communications, Inc., 881 F.3d 293 (4th Cir. 2018), is the cautionary case. Cox had a written thirteen-strike policy. It did not follow it. Internal records showed the company reinstating accounts it had just terminated and, in one message, describing the process in terms that made the policy look like theater. The Fourth Circuit held that a policy exists on paper only if it is reasonably implemented in fact, and Cox lost § 512(a) entirely. The Ninth Circuit had drawn the line earlier in Perfect 10, Inc. v. CCBill LLC, 488 F.3d 1102 (9th Cir. 2007), which held that a working system for receiving notices and acting on them, not a particular strike count, is what “reasonably implemented” demands.
What that loss costs, however, changed dramatically in 2026. In Cox Communications, Inc. v. Sony Music Entertainment, 607 U.S. ___ (2026), decided March 25, 2026, the Supreme Court ruled unanimously in Cox’s favor and reversed the Fourth Circuit’s contributory infringement holding, ending the liability that a $1 billion jury verdict had rested on. Justice Thomas wrote for seven justices; Justices Sotomayor and Jackson concurred in the judgment only. The Court held that a provider is contributorily liable only if it induced infringement or offered a service tailored for it, and that ordinary internet access has substantial lawful uses. Knowledge that some users infringe, plus a failure to terminate them, is not intent. The practical lesson is that losing the safe harbor and being liable are two separate questions, and Cox lost the first while winning the second.
What is the DMCA safe harbor? The DMCA safe harbor is 17 U.S.C. § 512, which shields a qualifying online service provider from monetary damages for copyright infringement committed by its users. It does not say the provider did nothing wrong. It says that if the provider meets the statutory conditions, the copyright owner cannot collect from it and must pursue the user instead.
What are the four DMCA safe harbors? Section 512(a) covers transitory digital network communications, meaning the conduit function an ISP performs when data passes through its pipes. Section 512(b) covers system caching. Section 512(c) covers storage of material at a user’s direction, which is the one hosting platforms rely on. Section 512(d) covers information location tools such as search engines and links. Each has its own conditions.
What does a platform have to do to qualify for safe harbor? Register a designated agent with the Copyright Office and renew that registration every three years, publish the agent’s contact information, adopt and reasonably implement a repeat infringer termination policy, inform users of it, accommodate standard technical measures, and act expeditiously on valid takedown notices. A provider must also lack disqualifying knowledge and must not receive a financial benefit directly attributable to infringement it can control.
How does a platform lose DMCA safe harbor? Most often by failing to reasonably implement a repeat infringer policy, which is what cost Cox Communications its § 512(a) protection in BMG Rights Management v. Cox. Other routes are actual knowledge of specific infringing items, red flag awareness of facts making infringement obvious, willful blindness, or profiting from infringement the provider had the right and ability to control.
Going further: How a DMCA takedown notice works, step by step .
This page is general legal information, not legal advice, and it does not create an attorney-client relationship.
The Ninth Circuit held volunteer moderators may be a platform's agents, so posts they screen may fall outside DMCA 512(c)'s user-direction safe harbor.
The Ninth Circuit held that defective DMCA notices cannot create knowledge, salacious site names are not red flags, and a tracked policy satisfies 512(i).
In a unanimous March 2026 judgment, the Supreme Court reversed the $1 billion verdict against Cox Communications, holding that knowledge alone cannot make an internet provider a contributory infringer, reshaping how the DMCA's § 512 safe harbor matters.
The Fifth Circuit held that a broadband provider can be contributorily liable for ignoring 1.3 million piracy notices, then vacated the $46.7 million award over how albums are counted. The Supreme Court vacated that judgment in April 2026 after Cox v. Sony.
The Ninth Circuit held that a copyright owner must consider fair use in good faith before sending a DMCA takedown notice, and that failing to do so can expose the sender to liability for misrepresentation under § 512(f).
A California district court held that an internet access provider is not a direct infringer for automatically copying user postings, seeding the volitional-conduct rule and the later DMCA safe harbors.
The Ninth Circuit held that Veoh's automated transcoding and playback functions fell within 'storage at the direction of a user,' and that general knowledge of infringement on a video platform does not defeat the DMCA safe harbor.
The Second Circuit held that both actual and 'red flag' knowledge under the DMCA must point to specific, identifiable infringements, but revived Viacom's case by reading willful blindness and 'right and ability to control' back into the statute.