Coders Lose Their DMCA Case Against GitHub Copilot and Codex
The Ninth Circuit has affirmed the dismissal of Digital Millennium Copyright Act (DMCA) claims against GitHub Copilot and OpenAI Codex, ruling that these tools create new work rather than simply stripping credit from existing copies. The court's decision, issued on September 18, 2026, rejected the argument that these AI models infringe copyright by reproducing code without attribution.
Key Rulings:
- No Removal or Alteration of Copyright Management Information: Judge Eric Miller stated that section 1202(b) of the DMCA requires the removal or alteration of existing copyright management information, which is not the case with Copilot as it generates new content.
- Comparison to Search Engines: The court compared Copilot to a search engine, emphasizing that search engines retrieve and display stored information, while Copilot generates predictions based on statistical patterns.
- Statutory Interpretation: The panel clarified that "identicality" is not a strict requirement under the DMCA, contrary to common perception. This interpretation will be crucial for future cases involving AI and copyright.
Background:
Anonymous programmers had sued GitHub, Microsoft, and OpenAI over Copilot and Codex, claiming these tools reproduce their open-source code without proper attribution, in violation of section 1202(b) of the DMCA. The panel dismissed this claim, leaving two breach-of-contract theories unaddressed.
The Programmers' Arguments:
- Output Theory: They argued that Copilot emits memorized training data without the necessary copyright notices.
- Input Theory (Abandoned): A claim that the defendants stripped copyright notices from code before feeding it into the training process, potentially breaching license agreements. However, this theory was dropped due to a statement by the programmers' counsel during a hearing.
Implications:
The ruling highlights the challenges of applying traditional copyright laws to AI-generated content. While the court acknowledged the awkwardness of its own prior cases regarding digital technologies, it provided clarity on the DMCA's application to AI.