Allow me to play devil’s advocate here, but what you are saying about the fediverse seems to be completely compliant with that license. The content can be freely redistributed provided it is fine in a noncommercial way and with attribution (which is the case, right? We see the comment author).
Also, the argument “X is going to be done regardless” applies to all licenses (thinking about open source licenses). There is nothing that physically stops you from taking open source code and violate its license but if you get caught doing so, you are liable.
Maybe today there is nothing that would make anybody accountable about grabbing public data, training AI on it and reselling it, but if in the future regulations will change, it will be hard(er?) for those companies to claim that certain content was distributed freely etc., in cases where the author explicitly and unequivocally stated the terms.
The content can be freely redistributed provided it is fine in a noncommercial way and with attribution (which is the case, right? We see the comment author).
There’s nothing preventing a Fediverse instance from showing ads, which would commercialize the comments on it.
Furthermore, they’re posting from Lemmy.world. Lemmy.world’s terms of service include this clause:
You waive Lemmy.World and its parent, subsidiaries, affiliates, and all their respective staff, representatives, service providers, contractors, licensors, licensees, and successors from any claims resulting from any action taken by Lemmy.World, and any of the foregoing parties relating to any investigations by either us or by law enforcement authorities.
That goes even further than the usual boilerplate on sites like Reddit that say “you grant us license to do whatever we want with the stuff you post here.”
And besides all that, copyleft licensing (and copyright in general) likely has no relevance to AI training regardless. Copyleft licensing only has power because it grants permission to make copies of something. You can actually reject a copyleft license, if you want, it just means that you can’t make copies of the thing once you’ve rejected the license. But training an AI doesn’t require making copies of anything, it only requires analyzing a copy that you already have. You don’t need permission to analyze something that you can already legally read.
There are of course some interesting court cases currently wending their way through various legal systems, and all sorts of legislation pending in all sorts of different countries, but as things stand right now that CC link is just pointless spam that’s being held up as a totem against witchcraft.
