There is no protection for calls to violence.
There is no protection for calls to violence.
I don't think anyone has asserted that Parler or its employees are advocating violence.
This is a violation of AWS's TOS which is why they're breaking it.
The comment I was responding to was misleading IMHO ("the speech Parler is being kicked off AWS for").
As you point out, they aren't being kicked off "for speech" they are being kicked off for insufficient moderation of someone else's speech, which speaks to the difficulties of being an at-scale content provider.
Again though, we’re getting lost in the weeds here. Parler’s entire selling point, and the only reason it had any users or any product, was the implicit promise that “we won’t moderate content”.
Maybe they were, it'll be tricky to show it for certain from our vantage point, but it certainly doesn't help their case that the whole point of platform that they weren't going to moderate said content.
You can’t build a platform for organizing murders, knowingly allow users to organize murders on your site, and then claim it’s not your fault because technically you aren’t responsible for UGC. Surely there’s a line where it’s just a veneer of an excuse.
Is it possible to build a platform with user generated content that is agnostic regarding the content in the same way that a traditional telco is agnostic about the content of phone calls (or SMS)? Is the difference that the messages are private rather than public?
We don't expect telco's to create filters or to independently moderate or to block the content of phone calls or SMS messages and basically legal mechanisms must be used to trigger any action against a subscriber. I realize that this is largely due to the legal framework of a "common carrier".
What I'm asking is, what would a "common carrier" look like for platform that had user generated content? Perhaps it isn't possible?
But perhaps there simply is an inherent problem with being a platform for user-generated content and that it requires somewhat robust moderation to avoid being cancelled, sued, or arrested.
I think you need to define was "common-carrier" means.
As far as I'm concerned, once a message is deemed to be intended to be public, "common carrier" goes away for the service that is going to be broadcasting that message to others. The copper/fiber/etc. lines between me and Twitter are still a common carrier, because even though my Tweet is going to be public, the individual TCP packets going to Twitter are still intended to be private.
Any system for publicly posting messages that isn't moderated will be overwhelmed by bad actors, since bad actors will nearly always be kicked off of moderated platforms.
The problem with a social service behaving that way is everyone who tried the "anything goes" approach over the last 30 years started idealistic either gave up and added some rules or became a toxic cesspool.
That makes sense to me. There are zero social environments on earth where some type of obnoxious behavior won't get you thrown out. Why would the web be all that different?
--
To be clear, this is a hypothetical about idealists acting in good faith. Over the years we've also seen bad actors feign utopian motives to try and avoid responsibility for their actions, which is a different conversation entirely.
100%. Absolutely.
> We don't expect telco's to create filters or to independently moderate or to block the content of phone calls or SMS messages and basically legal mechanisms must be used to trigger any action against a subscriber. I realize that this is largely due to the legal framework of a "common carrier".
Not only do we not expect it, but the more privacy-oriented people expect it to not even be possible, and use things like end-to-end encryption that make it impossible for the carrier to know what the messages contain.
There were guillotine memes with 50k upvotes on reddit right before the "insurrection" happened!