DOE vs. GitHub, INC: LLM generated-content not a DMCA violation [pdf] (cdn.ca9.uscourts.gov)

In a recent ruling by the United States Court of Appeals for the Ninth Circuit, the court upheld a lower court's dismissal of a Digital Millennium Copyright Act (DMCA) claim brought by a group of programmers against GitHub, Microsoft, and OpenAI. The plaintiffs alleged that GitHub's AI tools, Copilot and Codex, reproduced their open-source code without proper attribution, violating their copyright rights. However, the court determined that the AI-generated outputs do not qualify as altered copies of existing works but rather as new creations that never contained copyright management information (CMI). Therefore, the court found that the plaintiffs could not substantiate a DMCA claim under Section 1202(b). This ruling is significant for the AI and machine learning (ML) community as it sets a legal precedent regarding the use of AI in code generation and copyright considerations. The court’s decision reinforces the notion that outputs from large language models trained on publicly available data do not infringe on existing copyrights unless they replicate the original works identically. This interpretation could influence how AI developers and platforms navigate copyright issues, potentially shaping future innovations in AI-assisted development tools and the responsibilities tied to copyright when leveraging large datasets for training.