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

[Krzysztof Kozlowski <krzk@kernel.org>]

On 12/08/2025 18:12, Paul E. McKenney wrote:
> On Tue, Aug 12, 2025 at 06:03:36PM +0200, Krzysztof Kozlowski wrote:
>> On 12/08/2025 17:49, Steven Rostedt wrote:
>>> On Mon, 11 Aug 2025 18:28:52 -0400
>>> Sasha Levin <sashal@kernel.org> wrote:
>>>
>>>> We have the following in our docs:
>>>>
>>>>          code from contributors without a known identity or anonymous
>>>>          contributors will not be accepted. All contributors are required
>>>>          to "sign off" on their code
>>>>
>>>> Which requires a real, known, human identity behind the "Signed-off-by"
>>>> tag.
>>>
>>> I guess the real question is, if you have AI write the code, do you have
>>> the right to add your Signed-off-by to it? Especially if you don't know
>>> what that AI was trained on.
>>>
>>> Does the Signed-off-by mean if later on, we find that the AI used an
>>> patented algorithm, the one that added their SoB can be in legal trouble?
>>
>>
>> Maybe we should be very explicit about annotating AI-generated patches
>> and instead of (see workflows discussion [1]):
>>
>> 	Assisted-by: ....
>>
>> expect different tag, like:
>>
>> 	Legal-risk-by:
>>
>> or:
>> 	Legally-questionable-because-of:
>>
>> [1] https://lore.kernel.org/r/20250809234008.1540324-1-sashal@kernel.org/
> 
> If you have to add one of those last two tags, my carefully considered
> advice is to refrain from applying the patch.
> 
> Applying a patch containing the first tag might not be wise, either,
> depending on the outcome of a number of lawsuits currently in flight.
> Plus there are a lot of other countries that might choose to weigh in.

Yes, that's what I wanted to imply. At least person applying the patch
and then later sending in pull request to next maintainer could not use
excuse "I did not know, that Assisted-by causes legal risk".

Best regards,
Krzysztof