A federal judge has refused to dismiss YouTube creators’ claims that Runway AI bypassed the platform’s technical protections to obtain videos for AI training, allowing a proposed class action to proceed without deciding whether the company is liable.
U.S. District Judge Lewis A. Kaplan issued the 15-page opinion on October 30, 2026, in Ace Cam, Inc. v. Runway AI, Inc., No. 1:26-cv-01362-LAK, in the Southern District of New York.
The plaintiffs, Ace Cam, which does business as Random Golf Club, and Businessing LLC, allege that Runway circumvented YouTube’s access controls to download their videos for its generative-video models. Those claims remain allegations. At this stage, the court accepts well-pleaded factual allegations as true to determine whether they state a legally sufficient claim.
Public viewing did not defeat the claim
Runway argued that publicly viewable videos were not protected by access controls covered by Section 1201(a) of the Digital Millennium Copyright Act. It characterized the alleged restrictions as controls on copying rather than access.
Kaplan rejected that argument on the pleaded facts. The creators described a controlled playback system involving measures such as a rolling cipher, session-bound URLs and CAPTCHA challenges. He found plausible allegations that Runway bypassed at least one protection, and that the measures could qualify as access controls even though viewers could watch the videos publicly.
The judge also rejected the argument that creators had to install the protections themselves. The complaint plausibly alleged that they authorized YouTube to deploy protections for their work.
What this means for media businesses
For broadcasters, video podcasters and other creators distributing work on YouTube, the decision highlights a legal issue separate from whether copying material for AI training is fair use: how a developer obtains access to the underlying files.
This is a district-court pleading decision, not a trial verdict or a nationwide rule declaring all scraping unlawful. It neither certifies a class nor awards damages. It does not decide whether generated videos infringe the plaintiffs’ works or whether AI training generally qualifies as fair use.
The distinction also matters when following RNN’s coverage of the separate Suno litigation: training-data copying, allegedly bypassing technical protections and infringing outputs are different legal questions.
Source: October 30 memorandum opinion, Document 61. RNN reviewed the complete opinion, including its signed conclusion. The public docket mirror remains incomplete; no subsequent order changing this ruling was identified in the sources reviewed.