The Pulse
Ninth Circuit Limits DMCA Claims Against GitHub Copilot
The Ninth Circuit affirmed dismissal of a DMCA claim against GitHub, Microsoft and OpenAI, holding that the plaintiffs had not alleged removal of copyright-management information from an existing copy of their code.
AI.info Team ·
One claim survives as two
One of the three remaining claims in a major lawsuit over GitHub Copilot is now gone. On September 16, 2026, the U.S. Court of Appeals for the Ninth Circuit affirmed the dismissal of programmers’ claim that GitHub, Microsoft and OpenAI violated the Digital Millennium Copyright Act by producing code without the attribution, copyright notices and license terms attached to the original material.
The ruling leaves two breach-of-contract claims pending in the Northern District of California. It also gives AI companies a significant defense against one theory of copyright liability: a model’s failure to reproduce copyright management information is not automatically the same as removing that information from an existing copy.
The Ninth Circuit opinion resolves an interlocutory appeal in Doe v. GitHub, a case brought by anonymous programmers who published open-source code on GitHub. The plaintiffs alleged that Copilot and OpenAI’s Codex sometimes generated verbatim or near-verbatim versions of their code without the attribution and license information that accompanied the original repositories.
Why the DMCA theory failed
Section 1202(b) of the DMCA prohibits the intentional removal or alteration of copyright management information, or the distribution of works while knowing that such information has been removed or altered. The statute covers details such as an author’s name, a copyright notice and terms governing use of a work.
The plaintiffs argued that Copilot violated the statute when it reproduced code from its training material without the original CMI. The Ninth Circuit rejected that theory at the pleading stage because the allegations described Copilot as generating new code rather than retrieving a stored copy and stripping information from it.
“The complaint describes an AI tool that does not look up and reproduce stored work but rather creates new work.”
Eric D. Miller, Circuit Judge, U.S. Court of Appeals for the Ninth Circuit
Judge Eric D. Miller’s opinion says a new work that never contained CMI cannot be said to have had that information “removed” or “altered.” The court relied on the ordinary meaning of those verbs, which it said imply an affirmative act affecting information attached to an existing work.
The court rejects a simple “identical copy” rule
The district court dismissed the DMCA claim after concluding that Section 1202(b) requires an identical copy of the original work. The Ninth Circuit affirmed the dismissal but described “identicality” as an imperfect label rather than a separate statutory element.
Literal identity is not required in every case, the panel said. A work that substantially reproduces an existing protected work while omitting its CMI may support an inference that the information was removed. Minor cosmetic changes, the opinion says, would not necessarily protect a defendant that copied nearly all of a work and deleted its attribution.
That distinction matters for AI-generated code. The plaintiffs’ complaint described Copilot’s output as a “modified format,” “variation” or “functional equivalent” of licensed code. The panel held that those allegations did not show removal of CMI from a copy of the plaintiffs’ work. They instead described a system that learned statistical patterns from existing code and generated a new response.
Standing was not the problem
The Ninth Circuit did not dismiss the case because the programmers’ alleged injury was too speculative. The panel held that the complaint plausibly alleged a substantial risk that Copilot would reproduce code written by the named plaintiffs without attribution.
The complaint cited research asserting that large language models can emit memorized training data verbatim. It also pointed to GitHub’s duplicate-detection feature, which allows users to block suggestions matching public code snippets of at least 150 characters. The panel said those allegations were enough to establish standing at the motion-to-dismiss stage, while warning that the plaintiffs would need evidence of the risk if the case reached summary judgment.
That finding separates the case’s jurisdictional question from its statutory one. The programmers were allowed to bring the claim, but the theory they brought did not satisfy the DMCA as pleaded.
The forfeited training-data theory
The plaintiffs advanced a second argument on appeal: that GitHub, Microsoft and OpenAI removed CMI from source code during the training process before feeding the material into Copilot. The Ninth Circuit called that the “input” theory and declined to consider it because the plaintiffs had forfeited it.
The opinion says the district court repeatedly understood the case as concerning Copilot’s output rather than the preparation of training data. At one hearing, the plaintiffs’ counsel answered “Perhaps it doesn’t” when asked whether copying training data into Copilot violated open-source attribution requirements. The district court later stated that the plaintiffs had not alleged injury from the defendants’ use of licensed code as training data.
Because the plaintiffs did not correct that understanding in the district court, the appeals panel held that they could not introduce the input theory for the first time on appeal. The ruling therefore does not decide whether removing CMI during AI training could violate Section 1202(b).
Copyright and contract claims remain separate
The panel also declined to decide whether Copilot’s outputs infringe the programmers’ copyrights. Its opinion says generated code may still be substantially similar to existing code, but it expresses no view on whether that similarity would support a conventional copyright claim.
That reservation keeps the decision narrower than a ruling that treats all AI-generated code as legally new or all model outputs as protected from copyright litigation. The court addressed only whether the allegations fit the DMCA’s provisions governing the removal or alteration of CMI.
The two contract claims remain in the district court. Those claims concern obligations associated with open-source licenses rather than the specific DMCA theory rejected by the Ninth Circuit. The appellate decision therefore narrows one route for challenging Copilot’s handling of licensed code without ending the broader lawsuit.