Ninth Circuit Narrows DMCA Liability for AI Coding Tools: Doe v. GitHub Output Theory Fails, Open-Source License Claims Live
On September 16, 2026, the Ninth Circuit upheld the dismissal of the DMCA claims in Doe v. GitHub Inc. (No. 24-7700) — one of the first AI-training lawsuits — handing OpenAI and Microsoft a significant appellate win on the theory that Copilot and Codex strip copyright management information from licensed code. Per Reuters: "The 9th Circuit upheld the dismissal of the DMCA claims on Wednesday after finding Copilot and Codex do not remove copyright information from existing works, but instead generate new works that never contained such information."
The court's reasoning draws a bright line between DMCA CMI-removal claims and traditional infringement: "If substantial similarity was 'all it took' to violate the DMCA, the statute would 'supplant traditional copyright protections and subject defendants to potentially ruinous liability under the DMCA's enhanced statutory damages.'" Crucially, the panel expressly reserved the harder question — "it was not deciding whether similarities between AI-generated code and copyrighted code could support a copyright infringement claim."
U.S. District Judge Jon Tigar had dismissed the DMCA claim below but allowed the open-source licensing claims to proceed, and those continue. The case, filed in 2022 by two anonymous developers seeking to represent open-source code owners on GitHub, alleged Codex and Copilot were trained on repository code "in violation of a federal copyright law — the Digital Millennium Copyright Act (DMCA) — and open-source licensing terms."
What it means: This is the first appellate ruling in the AI-training litigation wave, and it forecloses one output-side liability theory (DMCA CMI) while leaving the two big questions untouched — whether training is fair use, and whether outputs that resemble training works infringe. For enterprises shipping AI coding assistants or code-generating features, the practical exposure now runs through (1) open-source license attribution/copyleft claims (still live in this case) and (2) traditional substantial-similarity infringement, which requires output-level evidence. Contract terms with AI vendors should address both indemnification for license violations and output-similarity claims. This fits the pattern in Ninth Circuit Narrows DMCA Liability for AI Coding Tools: Doe v. GitHub Output Theory Fails, Open-Source License Claims Live: courts are narrowing novel statutory theories against developers while core infringement theories remain unresolved.