Re: [MAINTAINERS SUMMIT] Annotating patches containing AI-assisted code

["H. Peter Anvin" <hpa@zytor.com>]

On August 11, 2025 3:12:25 PM PDT, "Paul E. McKenney" <paulmck@kernel.org> wrote:
>On Mon, Aug 11, 2025 at 02:57:30PM -0700, Luck, Tony wrote:
>> On Mon, Aug 11, 2025 at 02:46:11PM -0700, Paul E. McKenney wrote:
>> > On Fri, Aug 08, 2025 at 10:31:27AM +0200, Krzysztof Kozlowski wrote:
>> > > On 05/08/2025 19:50, Sasha Levin wrote:
>> > > With AI you do not have to even write it. It will hallucinate, create
>> > > some sort of C code and you just send it. No need to compile it even!
>> > 
>> > Completely agreed, and furthermore, depending on how that AI was
>> > trained, those using that AI's output might have some difficulty meeting
>> > the requirements of the second portion of clause (a) of Developer's
>> > Certificate of Origin (DCO) 1.1: "I have the right to submit it under
>> > the open source license indicated in the file".
>> 
>> Should the rules be:
>> 
>> 1) No submissions directly from an AI agent. The From: line must
>> always refer to a human.
>> 
>> 2) The human on the From: line takes full responsibilty for the
>> contents of the patch. If it is garbage, or broken in some way
>> there's no fall back on the "but AI wrote that bit".
>
>Another option is "The AI was trained only on input having a compatible
>license."  Which, to your point, would to the best of my knowledge cut
>out all of the popular and easily available AIs.
>
>There might well be less restrictive conditions on the AI training data,
>but I am not qualified to evaluate such conditions, let alone construct
>them.
>
>							Thanx, Paul
>

I think we need legal advice on this, but I think this is a *really* important issue. It could end up being a very ugly mess otherwise.