Hi Thanos and Greg, On 2026-05-10 at 07:54+03:00, Thanos Apollo wrote: > Ian Eure <[email protected]> writes: > > It is both process and artifact, as the one produces the other. > > If my process for creating software is to reuse as much infringing > > code as possible, the created software is tainted because of that, > > and its users bear some risk -- even if the risk is merely that > > they can no longer use it. > > By this logic any artifact produced through a process touching > copyrighted material is tainted. The English I'm writing this in was > learned from copyrighted books and films. Is this email tainted? At > some point process-purity arguments collapse under their own weight.
I understand this is taking the subject literally (if LLM is used), though be reminded that you writing an email is different from only pasting media quotes and expecting Springer and Disney not claiming rights to your messages. On 2026-05-10 at 07:54+03:00, Thanos Apollo wrote: > Ian Eure <[email protected]> writes: > > It matters because the provenence of the code is different. Prior > > to LLMs, "tool" meant, say, nano, or vi, or Emacs, or a > > typewriter, or some other thing which a human used to directly > > create the code. In that scenario, I agree, the tool does not > > matter. This situation is entirely different, as the provenance > > of the code no longer ends at the contributor, because they did > > not create it. Thus, there is a substantial risk of bugs due to > > inattention, or of the code not being copyrightable, or containing > > infringing material. > > Those risks exist for hand-written code too. Inattentive developers, > uncopyrightable trivial code, and infringing copy-paste predate LLMs. > Review is how we catch them, regardless of origin. I do not think it is beyond reason to draw a clear line at piping LLM directly into ones' editor and claim distribution rights over the codebase. A developer could learn from various sources, books, fora, web logs, or their aggregation/compression like LLMs. People are not arguing that one can copy from Stack Overflow and slap the Expat license on it. Why does it matter if someone indeed does that when the discussion is about reaching an ideal? I don't think we can catch such infringement through review either. Projects can ask nicely via contribution guidelines, and contributors can choose to be honest (or not). On 2026-05-06 at 07:26-04:00, Greg Hogan wrote: > On Tue, May 5, 2026 at 5:45 PM pinoaffe <[email protected]> wrote: > > And even if llm output is generally thought to be licensable, this > > clearly cannot apply to any near-perfect copies of some part of its > > training data that it may randomly emit, so incorporating llm output > > into a GPL project would likely still be a legal risk > > This is not happening in 2026. With old models and non-random > extraction, perhaps it can be done, but no one is demonstrating a > modern LLM returning "near-perfect copies of some part of its training > data" for any copyrightable unit of work. Just as with crypto where > important research is done on weakened algorithms (reduced iterations) > the demonstrations of targeted extraction and fine-tuning is reducing > our risk as mitigations are developed and applied. I would like to see studies backing this claim. Oracle published the interim policy for OpenJDK just last month, not last year. As for a demonstration try this classic prompt: > Complete the following: float Q_rsqrt Kind regards, Phong
signature.asc
Description: PGP signature
