A social network founded by a former OpenAI employee was caught importing public posts from Mastodon...and ran AI analysis to add tags to them.

Maven, a new social network backed by OpenAI’s Sam Altman, found itself in a controversy today when it imported a huge amount of posts and profiles from the Fediverse, and then ran AI analysis to alter the content.

  • misk@sopuli.xyzEnglish
    0·
    5 months ago

    That’s why I keep saying it’s pointless to defederate corpos. They’ll just scrape everything before you notice.

    • Blaze@reddthat.comEnglish
      0·
      5 months ago

      Defederation is more about not being flooded with 1000x more users than the Fediverse currently has

      • misk@sopuli.xyzEnglish
        0·
        5 months ago

        So far we only have a corpo fedi-twitter in form of Threads. In that case non-corpo instance user has to specifically follow someone before their content is federated so that sounds like a bit overblown issue.

        • Blaze@reddthat.comEnglish
          0·
          5 months ago

          Seems pretty easy for any corporation to setup something like https://lemmy-federate.com/ but for Maston/IceShrimp/Misskey accounts to federate the important corporate accounts to the targeted non-corpo instances

      • mke@lemmy.worldEnglish
        0·
        5 months ago

        And it’s also damming for private messaging on mastodon.

        I once read vague complaints about it being a rushed implementation. While I won’t trust those without evidence, I for sure wouldn’t trust mastodon with my PMs. At least, not until how this was allowed to happen is figured out and fixed if necessary.

        • technomad@slrpnk.netEnglish
          0·
          5 months ago

          I was under the impression that DM’s on Mastodon (and Lemmy too) weren’t ever stated as being secure and I think that they were both pretty transparent about this particular aspect.

          • mke@lemmy.worldEnglish
            0·
            5 months ago

            You’re right, regarding Mastodon. I won’t edit my other comment, though, both to preserve the original chain of thought and because that brings up another discussion.

            To quote the EFF:

            We feel that the intended usage of the feature will not determine people’s expectation of privacy while using it.

            Offering people a feature with preexisting expectations, similar to other things that fulfill those expectations, then telling people “We know it looks like a duck but don’t expect it to quack!”

            …It begs the question: was the feature really a good idea?