Re: Dealing with LLMs in IETF discussions draft

On Wed, 5 Aug 2026, George Michaelson wrote:

> I don't think we can just sit back and assume we're ok. I think we 
> should take an active position in this and declare requirements which 
> suit our needs.

Yes, I agree. If someone submits something under their name, they "own" 
it. But also, even when there is no AI involved, mistakes are made. A lot. 
AI and humans make mistakes. Often different mistakes, but still.

If I wrote something, used AI a bit to massage the text, I don't want to 
have to declare that AI was involved. This is the EU cookie law popup 
mistake in a different form. After a while it'll be noise, because 
everybody will do it. We should just accept that AI is here and it will be 
used by everyone, and just like the driver is responsible for what the car 
does, sometimes we also say "accidents happen" and say the driver isn't 
"criminally" at fault (which in IETF context means no rules were broken).

Not sure how it works in english publications, but in Swedish ones it's 
often "the car hit the cyclist" which always has been a pet peeve of mine. 
It's the driver of the car that hit the cyclist. Same thing here, it's the 
"driver" of the AI that is responsible.

-- 
Mikael Abrahamsson    email: [email protected]