The harm to AWS is in "interfer[ing] with AWS’s ability to prevent its services from being used to promote—and, as the events of January 6, 2021 have demonstrated, even cause—violence."

I wouldn't call that a laughable claim...

Plenty of people have posted similar or worse things on Twitter without moderation by Twitter or much shame to Twitter's upstream service providers.

Also, can you draw a direct causal line between AWS hosted Parler posts and actual violence that took place on Jan 6th?

I actually agreed with you until I read the order. It's about public good. Parler is arguing that Amazon is part of a cartel and that that's harmful to public, so the injunction restores fairness to the market. The judge finds that meritless because they have no evidence that the cartel exists, and so the injunction does nothing for the public good. I'm certain the factual errors about AWS hosting Twitter's feed didn't help.

Amazon argues that Parler had a role in the insurrection, so banning them is in the public good. Given the amount of content from the insurrection on Parler, it seems more likely than not that at least one person there was influenced by Parler. There is no requirement that absolute proof be shown, the judge is weighing probabilities here. It also helps Amazon that injunctive reliefs are considered an extraordinary measure, so Amazon wins by default unless Parler can reach a significant bar.

This was a moonshot, and Parler knew it. They're asking for a strong legal response based on a pretty flimsy case. Even if what they are saying is true (which I strongly doubt, but what if), they simply don't have the evidence they need to get this relief.

>it seems more likely than not that at least one person there was influenced by Parler.

I think that the real, concrete harm done to Parler's by their instant removal from Amazon is definitely not balanced by whatever "harm" in some Parler client's head before 6 Jan. I don't think it's a close call, and it is scary that federal judges are willing to make such a judgement.

The standard here seems to be that any unsubstantiated theoretical harm from user-uploaded content can be used as a basis to instantly kill any AWS client's business, if Amazon suddenly feels like doing so.

Probably†, but they don't have to. AWS's AUP sets the bar for what is or isn't acceptable on AWS, and AWS has 64 pages of Parler posts that contravened that AUP, many of which AWS warned Parler about and got no response.

† We'll all know soon enough, since Parler got dumped by a trivial IDOR vulnerability almost the moment AWS gave them notice they were going to be taken offline.

Yes, but apparently the work of drawing causal inference is unneeded because the judge already assumed it in her ruling, perhaps using her amazing oracular or exploit powers.