Basically the title. If someone can point me to a list of Lemmy and/or Piefed instances, that would be great! I’m looking for instances that forbid training AI on posts, use some kind of technology that makes AI training harder, or have at least promised to never train AI on users’ posts themselves. Or if you’re on an instance that has taken a stance like that, I’d love to hear about it.

The individual instance policy doesn't matter because anyone can spin up a Lemmy instance and get the data via federation bypassing any antibot crawling or instance policy.

you should just leave the internet, you'll be happier, AI isn't going anywhere anytime soon

While I suspect you are correct in your assumption about ai, there are better ways to convey this than telling someone to leave the Internet .

A bot scraping the fediverse would be indistinguishable from any other kind of federation, only the purpose differs.

Dang, I suspected that might be the case.

the instance's stance matters very little, for a lot of reasons:

1. nobody hosting a lemmy instance has the resources or interest to do ai training,
2. there isn't enough on lemmy alone to constitute a useful dataset,
3. nothing on lemmy is encrypted, save user credentials; every post and message you've ever sent is completely public. there's no way to make it harder to scrape without modifying the software, and even if you went that far,
4. in order for federation to work, everyone must be able to read everything. everything you post literally gets copied to other instances so they can see what you wrote.

the only way to be sure that nobody is using your instance to train ai is to run your own instance on a private network with federation turned off.

There are already instances doing this (and Mastodon ones as well), it's a thing - there are warehousing no-user instances[1] But they could be blocked! We can't have that, luckily...

...the Lemmy API and RSS feeds have no authorization around them for consuming the firehose of data; if I were a person doing this data gathering, I'd just passively monitor the firehoses off the top instances from random rotating IPs. Could and is most likely happening right now, can't block them.

[1] I personally wrote a simple script which uses the fediverse observer info, grabs the XX most popular communities and builds a subscription import file to bootstrap a new Lemmy instance. wasn't hard, just some curl and jq work.

I don't think it's a case of "allowing" ai crawlers to scrape data.

The AI data crawlers scrape everything, it's a matter of stopping them, not if they allow it explicitly. And these companies have found other ways around larger players that have tried to stop them.

Don't argue with sloppers. Just downvote, block, and move on.

sh.itjust.works

I don't know how true that actually is: sh.itjust.works just blocked like 3 networks they detected doing AI scraping.

is to run your own instance on a private network with federation turned off.

But I like talking to myself!

Bots scraping the website cause high levels of server load because compiling a page is an intensive activity, and so instance admins are taking active steps to stop this in order to stop all the scrapers bringing down the server.

But if they actually wanted Lemmy content (as opposed to the bots scraping everything on the internet they can find, and happening across Lemmy) then they can just start an instance and have other instances send them the content via normal federation. It would be almost impossible to tell this from a normal instance, and would be more efficient for everyone involved.

Well don't give them any ideas lol

As someone else already mentioned, it's almost certainly already happening. Lemmy federates with Mastodon, and Mastodon has a user base and a volume of content that would be hard to ignore. So there are likely many Mastodon instances that are slurping up content from Mastodon, Lemmy, Piefed, Pixelfed, anything else across the Fediverse, and using this to train AI.

It would just be nice if they didn't crash servers doing it.

Thanks!