--- /dev/null Thu Jan 01 00:00:00 1970 +0000
+++ b/hugo/content/post/facebook-republished/index.md Tue Aug 15 23:53:06 2023 -0500
@@ -0,0 +1,39 @@
+title: "Facebook Republished" +date: 2023-08-15T04:11:59-05:00 +This is a follow up to the previous post about +[Facebook taking down our page]({{<ref "post/facebook-takedown" >}}). It's a +bit late as I've been busy with other things, so apologies for that. +So on Wednesday August 9th I received an email from Facebook saying our page +has been published again! Hooray! +{{< figure alt="E-mail from Facebook saying our page has been published again." src="page-published.png" >}} +So we're all done right? Yes, but there's an interesting story here and we've +gotten the okay to talk about it, so lets do just that! +As you may have noticed, Gary tried to find help with this all over. His blog, +Facebook, X/Twitter, Mastodon, even LinkedIn among other methods. However, a +retired Pidgin developer, Ethan, reached out to a contact he had at Meta and +that's when things actually started happening. +His contact let us know that the page got taken down because Meta had just +started up a new AI model whose purpose is to help them find and take down +pages posing as Meta properties. Since the title of our page is `Pidgin Instant +Messenger` it got auto taken down by the AI due to the use of `Messenger` and +was never looked at by a human reviewer. +The contact went ahead and filed an internal appeal to the page being +unpublished. Once reviewed by a human, it was very clear that we were not +impersonating a Meta property, so Meta went ahead an republished the page. They +also added our page to their seed set/corpus so it should not be taken down +again in the future for impersonation. +Luckily this story had a happy ending, but it seems like that only happened +because we had contacts inside of Meta that could push things along. At any +rate, this is a good case for showing that AI is only as good as its input. \ No newline at end of file