The interlocutory appeal of the interpretation of Section 1202 of the DMCA is now being briefed. Represented by Joseph Saveri Law Firm and David Boies’s law firm, the Doe Appellants make the following arguments on appeal below. The defendant-appellees’ Answering Brief Due (Appellee) is due 5/12/2025. Optional Reply Brief due 21 days after service of Answering Brief. All briefs shall be served and filed pursuant to FRAP 31 and 9th Cir. R. 31-2.1.
One response to “Doe 1 v. Github 9th Circuit brief of Doe Appellant on DMCA CMI claims”
I can’t see the appellate court ruling contrary to plaintiffs’ base pleading. The statute does fall largely apart if it is to be strictly on identical copies.
But,
The argument that this applies to generative A.I. *outputs* – i.e. a generative A.I. generates code that is substantially similar to an existing work, and the omission of that work’s LICENSE declaration counts as removal of CMI – is problematic for other reasons. It first assumes that there was a version of this code that was output that had CMI, which was subsequently removed. Not so, as far as the majority of LLMs go.
Arguing that it was removed in training would be more appropriate, though I would argue that for the sake of the task, having – for example – the MIT license in the training data millions of times for each source code file otherwise bereft of that CMI would be extremely counterproductive as well.
A more appropriate way to handle this is to look for those substantial similarities, and label substantially similar code as such, pulling from the matched repository’s own LICENSE file to display potential licenses.
I do say potential licenses plural as there would be no shortage of surprised looks on the eyes of developers once they realize that an AI declares that a section of their own code licensed as MIT is substantially similar to a GPLv3 work from years earlier, and just like that their own repository is under licensing review.