Former Twitter employee details admins' extensive access to user data, personal messages

A screen shot in the Twitter Files drop from Bari Weiss on Thursday night revealed that Twitter employees have access to users' direct messages, and also showed a list of other tabs that are available internally available. 

These tabs confusing to the outsider were explained by Kristen Ruby, who reported that she spoke to a Twitter employee who gave her details on the use of those tabs.

Those tabs read: Timeline, Spaces, Live Videos, Communities, Guano, Email, Phone, Lists, Direct Messages, Moments, Interactions, Account Compromise, Spam, Payments, and Batch Action Tool.

Guano, Ruby explained with the help of an unnamed Twitter employee said that this tab contained databases. "So every database serves a purpose. Tracking actions or tweets etc etc. we had probably 100s if not 1,000s of databases each with separate tables."

This "raw data existed in guano table in each (most?) databases. They were called guano because they had 'dirty' non-formatted data. You could also make notes in them. If a tweet was 'act[i]oned' meaning we marked it in some way it would have a guano entry." 

Often "guano" was a collection of "notes on a tweet."

The "Testing" tab is 'when you have a machine learning algorithm and want to 'test' to see how good a job it did." She further explained that Twitter uses "any thing on the site" to test AI. "We used everything to train."

As to users' direct messages on the platform, the anon Twitter employee said that "Yes. We could read them because they were a source of abuse or threats. Again, strict protocols, but yeah. So if I am being harassed I might report that so Twitter can take action."

But the Twitter employee later said in the thread, "We had access to tweets. That's how we trained our models. We would mark a tweet as misinfo (or not) or abuse or whatever like copyright infringement. People who had access were pretty strictly controlled and you had to get sign off to get access. BUT a fair amount of people working in Trust and Safety did have it."

Privacy protocols, the Twitter employee said, were dealt with "by trying to restrict access to need to know."

"As far as the bans," the Twitter employee said, "The trends blacklist is indicating it's on a list of trends blacklisted. Meaning it violated one of our policies."

Twitter CEO Elon Musk confirmed that American political candidates were blacklisted while running for office. It was revealed that Libs of TikTok was intentionally banned by Twitter leadership, despite the account not actually being in violation of any rules, which was acknowledged by execs. Charlie Kirk and Dan Bongino were also suppressed.

"Protocol:" Ruby wrote, followed by a screen shot reading "Yeah. I don't know. That note up top is interesting. We couldn't even take action on accounts with over 10k followers. That had to be raised."

Shadowbanning, which was a big issue in the Twitter Files, also came up in Ruby's thread. The Twitter employee told her "I don't mean to sound duplicitous," when asked if Twitter engaged in the practice, "but let's first define shadow banning. Twitter considered shadow banning to be where the account is still active but the tweets are not able to be seen. The user is literally posting into thin air."

"Here is what we did," the Twitter employee continued. "We had terms (which, yes, were not publicly available) along the lines of hate speech, misinfo, etc. if your account repeatedly violated this but we decided not to ban you - maybe you were a big account, whatever - we could 'deamplify.' That's what I mean by corporate speech seeming shady. Lol. So, your tweet and account were still available to your followers. They could still see, but the tweets were hidden from search/discovery."

That employee said the platform "tried to be public about it," knowing that it was "Not a perfect system. You could argue, reasonably I think, that this is shadow banning. But that's why we said we didn't."