On Wed, 3 Jun 2026, Eric Rescorla wrote:
I'm sorry I wasn't clear. I agree that in the US AI-generated content cannot
be copyrighted. What I meant to challenge was that a disclosure of AI
text was required to "ensure that only the human-generated portions
are protected". This guidance seems to say that you need
to disclose when you register for copyright, but not generally.
Ah, we still agree. When you make a contribution, you have to be able to
grant the 5378 rights, but you don't have to provide details. In practice
contributions are usually a combination of previous contributions, public
domain, and stuff you wrote yourself. The only time I'm aware we ask for
more details is when material comes from a pre-5378 RFC and you might want
to get approval from the original authors.
I think we agree it would be polite to say if an LLM wrote your document,
but it doesn't matter for our process.
R's,
John
I would observe
that this problem already exists pre-AI because we might be incorporating
public domain material, but we don't designate that specially. Nor would
the kind of disclosure you propose be helpful unless we were to explicitly
designate which portions were unprotected.
Agreed. We haven't tried to mark public domain material in the past and I see no
need to do so now. What is or is not in the public domain is somewhat country
specific and it is not a rathole we need even to approach.
--
rswg mailing list -- [email protected]
To unsubscribe send an email to [email protected]
