On Wed Aug 5, 2026 at 4:36 PM CEST, Theodore Tso wrote:
On Wed, Aug 05, 2026 at 10:59:27AM -0500, Gerardo Ballabio wrote:
As I understand it, the problem with copyright isn't that AI providers
might claim copyright. It's that *someone else* might claim copyright
because the AI scraped and regurgitated their code. That's still an
open legal question AFAIK.
[...]
So it's a lot more complicated than that. For example, even if it is
a court decides that it wasn't fair use to scrape New York Times to
train the Perplexity's or Microsoft's LLM, it doesn't follow that the
New York Times could sue an open source project under a copyright
claim --- what is the likelihood that some of New York Times
intellectual property would end up in some project's C code if they
used Perplexity's LLM?
I think this wasn't the point at all. The issue may arise when the model
reproduces someone else's code in my codebase. Of course it's not going
to reproduce a newspaper in my C project.
