Key holding or allegation
The Ninth Circuit rejected a freestanding strict identicality rule but affirmed dismissal: on the pleaded output theory, newly generated code that never contained plaintiffs’ CMI was not a copy from which CMI had been removed. It left the input-stage training-data theory unresolved because plaintiffs had forfeited it below.
Why it matters
Leading U.S. appellate generative-code/CMI decision distinguishing AI-output §1202 theories from unresolved input-stage CMI removal.
Facts
Open-source developers challenged Copilot's use and output of code, including alleged removal or alteration of copyright-management information.
Issues before the court
Whether DMCA §1202 requires literal identity between source and challenged copy; whether AI outputs can be copies from which CMI was removed; and whether an input-stage CMI-removal theory was preserved.
What the court did not decide
The court did not decide ordinary copyright infringement based on substantial similarity, overall Copilot training legality, or the merits of the forfeited input-stage CMI-removal theory.
Procedural timeline
U.S. Court of Appeals for the Ninth Circuit
Interlocutory appeal opened: Interlocutory appeal concerns GitHub Copilot copyright-management-information issues, including the relationship between generated output and source work.
U.S. Court of Appeals for the Ninth Circuit
Published appellate opinion: Ninth Circuit affirmed dismissal of §1202 claims. It rejected a freestanding strict identicality rule, but held the pleaded output theory failed because newly generated code never containing plaintiffs’ CMI was not a copy from which CMI had been removed; input-stage removal theory was forfeited and left unresolved.