[rpd] The usage of AI in platforms

[JP]

On 22 Jul 2026, at 12:15, Mzwakhe Mabasa via RPD wrote:

> Good day.I would like to post to the RPD mailing list. Kindly allow me 
> this opportunity of being part of the RPD afrinic policy decisions. 
>
> While AI is widely used across the internet,  its acceptability 
> depends on the specific platform you are referring to . Most social 
> networks allow AI assistance, but require the clear labeling of 
> AI-generated imagery and ban non-consensual deepfakes or harmful 
> content. LSince platform policies vary greatly, here is a breakdown 
> of how the most popular services handle AI usage:
>
>    - Social Media (Facebook, Instagram, TikTok, YouTube): Generally 
> allowed, but social media algorithms penalize low-quality, generic 
> content. Major platforms are shifting to automatic detection and 
> require synthetic imagery to be appropriately labeled to avoid 
> misinformation penalties. 
>
>    - Design & Creative Platforms (Canva): AI usage for brainstorming 
> and generating designs is permitted, but platforms typically place 
> operational limits on how much AI content you can generate per month. 
> Users are responsible for ensuring AI outputs do not mislead others 
> and are used in accordance with local copyright laws. 
>
>    - AI Generation Models (Open AI, Midjourney): AI tools are 
> permitted, but they enforce strict universal policies. You cannot use 
> these services for harassment, defamation, or generating unauthorized 
> deepfakes of real people

to play off an older joke:

q: “how do you know someone’s using LLMs?”
a: “they’ll damn well tell you”

normally I don’t need to pay much attention to rpd because things 
don’t move that much, and then the last week… the sloppist tsunami 
has been quite notable. and it’s near all this kind of milquetoast 
nonsense gibberish. I saw someone else mention the astroturfing 
component of things too (and a cursory investigation around the poster 
domains for the most vocal proponents there had me raising my eyebrows 
most spock-like) but I’ll leave that aside for now

it’s probably worth rpd considering a full and complete ban on 
prompt-generation and prompt-assistance for use on the policy list, with 
possibly a poster cooldown/block period (and some way to deal with 
repeat offenders and ban evasion).

there’s both precedent and strong arguments in favour

precedent: a number of other public-benefit and mass-cooperation 
organisations and projects have taken to the hardline block (the most 
notable is codeberg, as of earlier today)

arguments: the most obvious is exactly what has happened in the last 
week (and the alternative of no ban is an utter deluge of worthless 
noise, which will space-fill as far as it can go). there is also ample 
evidence across multiple years and multiple studies that these 
technologies are _bad_ at fine detail, and that’s exactly the sort of 
thing that is very ill-fitting in the context of what rpd is to be doing

-J
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.afrinic.net/pipermail/rpd/attachments/20260722/f83dfe25/attachment.html>