Ninth Circuit Affirms Dismissal of DMCA Claims Against GitHub and OpenAI
Bot Mutiny |
The U.S. Court of Appeals for the Ninth Circuit ruled that AI tools Copilot and Codex do not violate the Digital Millennium Copyright Act because they generate new works rather than altering existing code copies.
The U.S. Court of Appeals for the Ninth Circuit has affirmed the dismissal of Digital Millennium Copyright Act claims brought against GitHub, Microsoft, and OpenAI. The decision, filed September 16, 2026, centers on how generative AI models handle copyright management information when producing code. According to documents published by CourtListener, a three-judge panel led by Judge Eric D. Miller concluded that the programmers who sued the tech giants failed to state a claim under Section 1202(b) of the DMCA. The court found that the AI tools in question, Copilot and Codex, do not actually remove or alter data from protected works. They just create new ones. ## The Output Theory
The plaintiffs are a group of programmers who published open-source code on GitHub under licenses that require attribution. They alleged that Copilot, which was trained on billions of lines of public code, reproduces their work "essentially verbatim" without including the required author names or copyright notices. The court rejected this "output" theory. The judges ruled that the DMCA provision specifically targets the removal or alteration of copyright management information from a "copy" of a work. Under the court's interpretation, the AI doesn't take an existing file and strip the credits. Instead, the model identifies statistical patterns to generate a new completion for a prompt. Because the AI's output is technically a "new work" that never contained the attribution to begin with, the court held that no information was removed or altered in the legal sense. ## Training and Standing
GitHub is the largest hosting service for open-source software in the world. Its Copilot tool, developed with OpenAI, uses large language models trained on "all available public GitHub repositories." The plaintiffs argued that as the model scales, the risk of it producing identical copies of training code increases. The Ninth Circuit did grant the plaintiffs one procedural victory by acknowledging they had Article III standing. The panel found the programmers plausibly alleged a "substantial risk of injury" regarding their code being used. However, standing isn't the same as a valid claim. ## Forfeited Arguments
The court also declined to hear the plaintiffs' "input" theory. This argument would have focused on whether the act of feeding the code into the model during the training phase violated the DMCA. The panel ruled this theory was forfeited for the purposes of the appeal. The ruling leaves the remainder of the case to focus on breach of contract claims. While the programmers proved they have a right to be in court, the specific protections of the DMCA have been ruled inapplicable to the way these models generate code. The panel consisted of Circuit Judges Sidney R. Thomas and Eric D. Miller, along with District Judge Stanley Blumenfeld, Jr., sitting by designation. The case is Doe v. GitHub, Inc., No. 24-7700.