DMCA § 512(c) Safe Harbor: Hash-Matching Outputs and Algorithmic Curation Do Not Create “Red Flag” Knowledge or “Right and Ability to Control” Without Specific Infringement and Substantial Influence

Athos Overseas Limited Corp. v. YouTube, Inc.
Court of Appeals for the Eleventh Circuit — January 7, 2026 (Jordan, J.)

1. Introduction

This Eleventh Circuit decision addresses how the Digital Millennium Copyright Act (“DMCA”) safe harbor for user-directed storage, 17 U.S.C. § 512(c), applies to a modern, large-scale video platform. Athos Overseas Limited Corp. (“Athos”), owner of copyrights in classic Mexican and Latin American films, sued YouTube (YouTube, Inc., YouTube, LLC, and Google, LLC) for copyright infringement based on user uploads. Athos did not meaningfully dispute that YouTube removes content identified by compliant takedown notices. Instead, Athos sought to defeat § 512(c) protection by arguing:

• Knowledge theory: YouTube’s hashing/matching technologies allegedly give it actual or “red flag” knowledge of additional infringing videos beyond the URLs identified in notices, and YouTube allegedly fails to remove those additional items.
• Control/benefit theory: YouTube’s moderation, curation, and platform controls (including algorithmic suggestion/autoplay) allegedly give it the “right and ability to control” infringing activity while deriving a direct financial benefit.

The district court granted summary judgment to YouTube on § 512(c) safe-harbor grounds; the Eleventh Circuit affirmed.

2. Summary of the Opinion

3. Analysis

3.1. Precedents Cited

Viacom Int'l Inc. v. YouTube, Inc., 676 F.3d 19 (2d Cir. 2012)

The court relied on Viacom for the framework distinguishing: (i) actual knowledge (subjective knowledge of specific infringement) and (ii) red flag knowledge (subjective awareness of facts that would make specific infringement objectively obvious). The Eleventh Circuit accepted Viacom’s core operational point: § 512(c) is structured around the capacity to remove material “expeditiously,” which presupposes knowledge “with particularity” about what to remove. This “specificity” requirement undercut Athos’ effort to convert generalized awareness (or tool-generated “possible matches”) into disqualifying knowledge.

UMG Recordings, Inc. v. Shelter Cap. Partners LLC, 718 F.3d 1006 (9th Cir. 2013)

The Eleventh Circuit aligned with UMG’s conclusion that general knowledge that a service can be used for infringement—without awareness of specific infringing items—is insufficient for § 512(c)(1)(A) actual or red-flag knowledge. The opinion uses UMG to reinforce that safe-harbor eligibility is preserved where the provider promptly removes specifically identified infringing material.

Capitol Recs., LLC v. Vimeo, Inc., 125 F.4th 409 (2nd Cir. 2025) (Vimeo II) and Capitol Recs., LLC v. Vimeo, LLC, 826 F.3d 78 (2d Cir. 2016) (Vimeo I)

These decisions were central to the “right and ability to control” analysis. The Eleventh Circuit adopted Vimeo II’s “something more” requirement: a provider must exert substantial influence on user activity or induce infringement; merely having the ability to remove content, set rules, or promote content is not enough. The court also invoked Vimeo I to support the proposition that § 512(c) functions as an affirmative defense and to frame willful blindness as compatible with § 512(m) only in limited, appropriate circumstances.

Lenz v. Universal Music Corp., 815 F.3d 1145 (9th Cir. 2016) and Perfect 10, Inc. v. CCBill LLC, 488 F.3d 1102 (9th Cir. 2007)

The Eleventh Circuit cited these cases for the content requirements of a valid DMCA takedown notice, emphasizing that the notice-and-takedown regime hinges on identification of the allegedly infringing material and information reasonably sufficient to locate it. This reinforced the court’s rejection of Athos’ attempt to shift to YouTube the burden to find additional matches beyond the noticed URLs.

Other cited authorities shaping the analysis

• In re Aimster Copyright Litig., 334 F.3d 643 (7th Cir. 2003): referenced to illustrate that safe harbors demand reasonable steps regarding repeat infringers, not a broad monitoring obligation.
• EMI Christian Music Group, Inc. v. MP3tunes, LLC, 844 F.3d 79 (2d Cir. 2016) and Capitol Records, Inc, v. MP3tunes, LLC, 821 F. Supp. 2d 627 (S.D.N.Y. 2011): used to underscore that when investigation is required to determine infringement, the circumstances are not “red flags,” and that § 512(m) limits any implied monitoring duty.
• In re subpoena of Internet Subscribers of Cox Commc'ns, LLC, 148 F.4th 1056 (9th Cir. 2025): cited for the function-specific nature of § 512 safe harbors.
• BWP Media USA, Inc. v. Clarity Digit. Grp., LLC, 820 F.3d 1175 (10th Cir. 2016): cited for the notion that automated processing typically does not defeat safe-harbor protection.
• Luxottica Grp., S.p.A. v. Airport Mini Mall, LLC, 932 F.3d 1303 (11th Cir. 2019): supplied a willful-blindness definition (imported here by analogy, consistent with Viacom).

3.2. Legal Reasoning

(A) The court’s factual gatekeeping: what the technology did—and did not—do

A notable feature of the opinion is its insistence on grounding DMCA knowledge and control arguments in an accurate technological record. Athos argued that YouTube “automatically” runs Content ID and match processes across uploads, such that a single takedown notice should effectively reveal and obligate removal of all matching copies. The court rejected that characterization as unsupported by the cited deposition testimony.

The court found, for purposes of summary judgment, that YouTube’s hashing and matching tools:

• generate hashes for uploads and can compare reference hashes to detect potential matches;
• are embedded in multiple tools (Content ID, Copyright Match Tool, and a “Prevent Copies” functionality);
• do not themselves determine whether a match is legally infringing (e.g., fair use, license, authorized uploads); and
• require substantial copyright-owner input and management, with the most powerful tools (Content ID, Copyright Match Tool) not universally available to all users.

This factual framing drove the legal result: outputs identifying “potential matches” do not equate to knowledge of “specific infringement,” and the DMCA cannot be read to force a provider to operate owner-facing tools on owners’ behalf to hunt down additional instances beyond notice.

(B) Actual knowledge and “red flag” knowledge under § 512(c)(1)(A)

The Eleventh Circuit concluded that Athos failed to raise a triable issue that YouTube had actual knowledge of infringing material it did not remove. Athos effectively conceded YouTube removes specifically identified URLs in compliant notices.

On red-flag knowledge, the court emphasized two interlocking statutory constraints:

• Specificity and expeditious removal: Under the operational logic endorsed in Viacom and UMG, the obligation to remove material “expeditiously” presupposes particularized knowledge of what to remove.
• No monitoring duty: § 512(m)(1) expressly provides that safe-harbor protection is not conditioned on monitoring or affirmatively seeking facts indicating infringement.

Athos’ principal “red flag” theory—accumulation of many takedown notices plus YouTube’s refusal to remove additional, non-noticed matches—failed because (as the court noted, invoking Nimmer as well) “fully meet[ing]” safe-harbor obligations by removing “the precise material noticed” does not convert a service provider into a general-purpose infringement investigator.

(C) The court’s treatment of Nimmer’s critique

Athos challenged reliance on Viacom and UMG by pointing to commentary suggesting red-flag knowledge may be “looser.” The Eleventh Circuit engaged that critique but ultimately held it unnecessary to resolve, because Athos’ argument rested on takedown volume and a demand for non-noticed removals—an approach Nimmer himself characterizes as inconsistent with the DMCA’s notice structure.

The court also expressed skepticism that a “generalities” approach to red-flag knowledge could be applied to a video-hosting platform without effectively imposing a monitoring duty barred by § 512(m)(1), especially where determining infringement often requires legal judgments.

(D) Willful blindness

The Eleventh Circuit agreed with Viacom that willful blindness can, in appropriate circumstances, show knowledge or awareness of specific infringement without nullifying § 512(m)(1). But it found no willful blindness on this record because:

• the tools at issue produced at most “lists of potential infringement,” not obvious infringement; and
• requiring YouTube to investigate those outputs to decide what is infringing would contravene § 512(m)(1)’s prohibition on requiring affirmative fact-seeking.

(E) “Right and ability to control” under § 512(c)(1)(B)

The opinion’s second major doctrinal contribution is its adoption of Vimeo II’s understanding that the “right and ability to control” requires “something more” than the ability to remove content or enforce policies. The Eleventh Circuit held that YouTube’s:

• ability to remove videos/users,
• setting of content rules, and
• algorithmic suggestion and autoplay

are “noncoercive exercises of control” and do not show “substantial influence” or inducement. Because Athos failed the “control” prong, the court did not reach whether YouTube receives a “direct financial benefit” from infringing activity.

3.3. Impact

• Eleventh Circuit alignment with major circuits on § 512(c): The decision brings the Eleventh Circuit into clear alignment with the Second and Ninth Circuits on (i) the need for particularized knowledge tied to removal obligations and (ii) a narrow construction of “right and ability to control” requiring “something more.”
• Limits on “hash match = knowledge” arguments: The court signals that automated hashing/matching outputs—without embedded legal adjudication (fair use, authorization, licensing)—are insufficient to impute actual knowledge of infringement and cannot be used to force providers into de facto monitoring.
• Reinforcement of notice design and § 512(m): The opinion reads § 512(c) as a system that relies on the copyright owner to identify and locate specific items, while § 512(m) prevents courts from converting technological capability into a legal duty to search.
• Algorithmic curation not “substantial influence” absent more: By treating suggestion/autoplay and general moderation powers as insufficient for § 512(c)(1)(B), the court reduces the likelihood that ordinary recommendation systems alone will defeat safe harbor, absent evidence of inducement or coercive shaping of user infringement.
• Record-limited caution: The court expressly “express[ed] no views on the application of § 512(c) in a case where a service provider has different functionalities and technologies,” leaving room for future litigation where the platform’s tools are more automated, more provider-directed, or more legally discriminating.

4. Complex Concepts Simplified

• § 512(c) safe harbor: A statutory shield that can protect a platform from copyright liability for user-uploaded material, if it satisfies conditions (including responding to proper takedown notices).
• Actual knowledge vs. red-flag knowledge: “Actual” means the provider actually knows specific content is infringing; “red flag” means the provider is aware of facts that would make specific infringement obvious to a reasonable person.
• Notice-and-takedown: The DMCA process where copyright owners send a compliant notice identifying infringing material with enough information (often a URL) for the provider to locate it; the provider must then remove/disable access expeditiously.
• § 512(m) no-monitoring rule: The DMCA does not require a provider to monitor its service or affirmatively look for infringement as a condition of safe-harbor protection.
• Hashing / hash matching: A technical method to create an identifying signature for a video and compare it against reference signatures to find “matches.” The opinion stresses that a “match” is not the same as legal “infringement” because authorization and defenses (like fair use) can apply.
• Willful blindness: A doctrine where someone suspects wrongdoing and deliberately avoids confirming it. Here, the court held YouTube was not willfully blind because the tools produced at most potential matches, and the DMCA does not require the provider to investigate to make legal judgments.
• “Right and ability to control”: More than the power to delete content; it requires “something more,” such as substantial influence over user conduct or inducement of infringement.

5. Conclusion

Athos Overseas Limited Corp. v. YouTube, Inc. solidifies, in the Eleventh Circuit, a DMCA § 512(c) approach consistent with Viacom Int'l Inc. v. YouTube, Inc., UMG Recordings, Inc. v. Shelter Cap. Partners LLC, and Capitol Recs., LLC v. Vimeo, Inc. (Vimeo II): safe-harbor protection is not lost merely because a platform has powerful technical tools or recommendation features. Absent evidence that the provider has specific knowledge of infringing items it fails to remove, is willfully blind to obvious infringement, or exerts “substantial influence” (or inducement) amounting to the statutory “right and ability to control,” § 512(c) remains available. The court’s analysis underscores that recalibrating this allocation of burdens—between copyright owners’ identification duties and platforms’ takedown obligations—is a task for Congress, not judicial revision.