Re: AGENTS.md

--000000000000493b89065839a382
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Tue, Aug 4, 2026 at 6:23=E2=80=AFAM Sean Whitton <[email protected]=
me>
wrote:

> Dr. Arne Babenhauserheide [04/Aug 10:23am +02] wrote:
> > Sean Whitton <[email protected]> writes:
> >
> >> But it does suggest that we would get at least some benefit from
> >> committing AGENTS.md and CLAUDE.md along the lines proposed in this
> >> thread?
> >
> > There=E2=80=99s also potential damage: when I see an AGENTS.md or CLAUD=
E.md file
> > in a repository, I usually decide that my time is better spent elsewher=
e.
> >
> > The presence of these files hints at accepting LLM contributions to me.
> >
> > If this is similar for many others, we may lose potential human
> > contributors who are aligned with our policy of not accepting LLM
> > authored code.
>
> This policy is only provisional.  We do not have a general anti-LLM
> stance.  Probably the FSF are going to allow LLMs that meet certain
> criteria of freeness.  So we will probably want an AGENTS.md anyway.
>

In case anyone missed it, GCC has a newly minted AI policy which is
presumably FSF sponsored https://gcc.gnu.org/ai-policy.html

I find it curious that their assumption is that the copyright for LLM
content attaches to its originating user and hence can be assigned to the
FSF.  In the US, LLM outputs are not copyrightable (automatically or easily
as "sufficient human contribution" is provably necessary).  On that basis,
the GCC policy does not read as if they consulted lawyers who read and
understand US Copyright Office guidance.  See V. Conclusion on page 41
https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-C=
opyrightability-Report.pdf

-St=C3=A9phane

--000000000000493b89065839a382
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div dir=3D"ltr"><div class=3D"gmail_default" style=3D"fon=
t-family:monospace"><span style=3D"font-family:Arial,Helvetica,sans-serif;b=
ackground-color:transparent">On Tue, Aug 4, 2026 at 6:23=E2=80=AFAM Sean Wh=
itton &lt;<a href=3D"mailto:[email protected]">[email protected]=
ame</a>&gt; wrote:</span></div></div><div class=3D"gmail_quote gmail_quote_=
container"><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.=
8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Dr. Arne Baben=
hauserheide [04/Aug 10:23am +02] wrote:<br>
&gt; Sean Whitton &lt;<a href=3D"mailto:[email protected]" target=3D=
"_blank">[email protected]</a>&gt; writes:<br>
&gt;<br>
&gt;&gt; But it does suggest that we would get at least some benefit from<b=
r>
&gt;&gt; committing AGENTS.md and CLAUDE.md along the lines proposed in thi=
s<br>
&gt;&gt; thread?<br>
&gt;<br>
&gt; There=E2=80=99s also potential damage: when I see an AGENTS.md or CLAU=
DE.md file<br>
&gt; in a repository, I usually decide that my time is better spent elsewhe=
re.<br>
&gt;<br>
&gt; The presence of these files hints at accepting LLM contributions to me=
.<br>
&gt;<br>
&gt; If this is similar for many others, we may lose potential human<br>
&gt; contributors who are aligned with our policy of not accepting LLM<br>
&gt; authored code.<br>
<br>
This policy is only provisional.=C2=A0 We do not have a general anti-LLM<br=
>
stance.=C2=A0 Probably the FSF are going to allow LLMs that meet certain<br=
>
criteria of freeness.=C2=A0 So we will probably want an AGENTS.md anyway.<b=
r></blockquote><div><br></div><div class=3D"gmail_default" style=3D"font-fa=
mily:monospace">In case anyone=C2=A0missed it, GCC has a newly minted AI po=
licy which is presumably FSF sponsored=C2=A0<a href=3D"https://gcc.gnu.org/=
ai-policy.html">https://gcc.gnu.org/ai-policy.html</a></div><div class=3D"g=
mail_default" style=3D"font-family:monospace"><br></div><div class=3D"gmail=
_default" style=3D"font-family:monospace">I find it curious that their assu=
mption is that the copyright for LLM content attaches to its originating us=
er and hence can be assigned to the FSF.=C2=A0 In the US, LLM outputs are n=
ot copyrightable (automatically or easily as &quot;sufficient=C2=A0human co=
ntribution&quot; is provably necessary).=C2=A0 On that basis, the GCC polic=
y does not read as if they consulted lawyers who read and understand US Cop=
yright Office guidance.=C2=A0 See V. Conclusion on page 41=C2=A0<a href=3D"=
https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-C=
opyrightability-Report.pdf">https://www.copyright.gov/ai/Copyright-and-Arti=
ficial-Intelligence-Part-2-Copyrightability-Report.pdf</a></div><div class=
=3D"gmail_default" style=3D"font-family:monospace"><br></div><div class=3D"=
gmail_default" style=3D"font-family:monospace">-St=C3=A9phane=C2=A0</div></=
div></div>

--000000000000493b89065839a382--