Re: Participation using AI in WG mailing lists and github repos

[Mark Nottingham <[e-mail]>]

I've recently instructed Claude to do this:

> * When you create a PR or comment, clearly indicate that the content is AI-generated, and briefly indicate the level of human supervision/interaction/awareness that led to it.

I like the results so far. Of course, someone can omit that -- either intentionally or not -- but I find it helps to keep me honest and assures that there's a good match between my use of AI and readers' understanding of it.

Cheers,


> On 9 Jun 2026, at 2:31 pm, Martin Thomson <[e-mail]> wrote:
> 
> On Tue, Jun 9, 2026, at 03:24, Tim Bray wrote:
>> On Jun 6, 2026 at 11:12:34 PM, Rohan Mahy <[e-mail]> wrote:
>>> **
>> **
>>> **Disclosure**
>>> Every new mailing list post on a new topic, and each new github PR or Issue should include the following disclosure and assertion:
>>> "This contribution was originally generated by <Model>[<version>]. I personally reviewed every word."
>> 
>> Why on earth would anyone care which version of which LLM was upstream? 
>> s/by <Model>[<version>]/software/.
> 
> At risk of misinterpreting Tim's point here, this is key.  We should not care how the contribution was generated, only that the full implications of that contribution properly attach to the reputation of the individual.
> 
> From my perspective, that means a *prohibition* on disclosures like this.
> 
> We could pretend reviewing is the same as writing, but it isn't.  Sure, we are able allocate reputation (positive or negative) to the individual who is willing to pass off the work of a bot as theirs. A disclaimer allow the individual to disavow any contribution as a reviewing error.
> 
> As for the policy, this is going to suck for a bit.  Because I'm seeing as much apparent slop as Mark.  We'll need to develop defenses.  But we should develop those with the expectation that there is a human on the other end.  Because what seems like slop to me is often still the earnest contribution of a human who might genuinely stand by the outcome.

--
Mark Nottingham   https://www.mnot.net/