OpenAI said in a Friday that the coders, who all sued anonymously, were asking the court to adopt an interpretation of the Digital Millennium Copyright Act that would give any copyright owner a DMCA claim.
The software developers sued OpenAI, Microsoft and subsidiary GitHub in 2022, claiming the generative AI Copilot and the LLM it was trained on, Codex, were trained on source code stolen from public GitHub repositories.
According to OpenAI, U.S. District Judge Jon S. Tigar joined a consensus when he held that the coders had not shown that OpenAI and Microsoft removed copyright management information from their work when he dismissed the DMCA claims.
OpenAI said the coders' had only alleged that Codex and Copilot could potentially generate their code to users without including information from their open source licenses, calling their alleged injuries "hypothetical and manufactured."
"The injury is conjectural and hypothetical because plaintiffs have never alleged that ordinary users would ever, much less are likely to, enter prompts into Copilot or Codex that cause the tools to generate variations of plaintiffs' code," OpenAI said. "Instead, plaintiffs generally allege that, 1% of the time, Copilot outputs 150-character snippets that match someone's existing code. But plaintiffs do not explain how that allegation applies to their code."
The only examples the coders gave of Copilot generating similar code was when the coders themselves prompted the AI with verbatim lines of their own code, OpenAI said.
OpenAI said the coders had changed course from their focus in district court on Copilot's and Codex's alleged output of their code to a new theory on appeal that their copyright management information had been removed from their code. But that "late-stage theory" should be rejected since it lacks support in their suit in district court, OpenAI said.
Microsoft said in a , also filed Friday, that the coders had no standing to sue because they "fail to allege that Copilot has or will output their copyrighted code."
"Though they allege their code is among the billions of lines defendants OpenAI used to train Codex — the LLM underlying Copilot — they offer no reason Copilot would register their code as significant or suggest it as output," Microsoft said.
Microsoft said the "emptiness" of the coders' claims was "most stark" in that they had not alleged that the removal of copyright management information was likely to lead to copyright infringement.
The coders had not alleged a plausible infringement of their code or explained how the use of short pieces of code from Copilot would significantly infringe a copyright, the brief said.
Counsel for the coders did not immediately respond Monday to requests for comment.
The coders are represented by Daniel Morales, Joshua Stein, Evan Ezray, Maxwell Pritt, Jesse Panuccio and Margaux Poueymirou of Boies Schiller Flexner LLP and Joseph Saveri, Evan Creutz, Christopher Young and William Castillo Guardado of Joseph Saveri Law Firm LLP.
OpenAI is represented by Joseph Gratz and Tiffany Cheung of Morrison Foerster LLP and Lisa Blatt, Thomas Hentoff, Claire Cahill, Erin Sielaff and Claire Lazar of Williams & Connolly LLP.
GitHub and Microsoft are represented by Annette Hurst, Nicholas Gonzalez, Christopher Cariello, Jonas Wang, Alyssa Caridis and Geoffrey Shaw of Orrick Herrington & Sutcliffe LLP.
The case is Doe et al. v. GitHub Inc. et al., case number 24-7700, in the U.S. Court of Appeals for the Ninth Circuit.
--Additional reporting by Ivan Moreno. Editing by Rich Mills.
For a reprint of this article, please contact reprints@law360.com.