It's September 2024, and a swarm of spammy X accounts is spamming people's replies with oddly repetitive attacks on Democratic U.S. presidential candidate Kamala Harris.
cc:
@ZellaQuixote
Popular chatbot ChatGPT will sometimes return error messages when asked to produce offensive content. If you're a spam network operator who uses ChatGPT to generate spam and you're not paying attention, these messages will show up in the spam you generate.
cc:
@ZellaQuixote
This video show the network requests that Twitter is currently making when I attempt to view a tweet. Just gonna go out on a limb here and speculate that maybe the actual problem is Twitter trying to refresh itself 20 times every second, rather than "scraping".
It's a Thursday, and a variety of Twitter accounts with GAN-generated faces have paid $8 for a blue checkmark.
#BlueCheckShenaniGANS
(GAN = "generative adversarial network", the AI technology used by tools like )
cc:
@ZellaQuixote
Spoiler alert: if you call yourself a "nationalist" and are totes ok with a foreign government spiking the punch in our elections just because you like the result. . . you're not actually a nationalist.
Not long after Julian Assange's arrest,
@seanhannity
deleted all of his tweets referencing Assange or
@wikileaks
. Thankfully, the Internet is forever and some of Sean's best work on the topic lives on in the form of retweets from banned Kremlin troll accounts.
As Western democracy continues to fray due in part to the willingness of folks like Julian Assange to play useful idiot and spread disinformation on behalf of Vladimir Putin, it's worth reflecting on why the actual fuck anyone listens to a rape suspect hiding in an embassy.
As Western power structures continue to fray due to the post-primacy of the US empire & dislocating effects of the internet, which has liberated cities & regions to deal with each other directly, are we seeing the invention of foreign & domestic enemies to bolster the old order?
I get it now. Sex crimes against children are a big deal when reported in Russia Today, but when a Republican Senate candidate endorsed by Nigel Farage and Steve Bannon is accused, he gets a free pass.
#AltWankers
#GoFarageYourself
.
Hospitals don't generally get a ton of engagement on Twitter. That all changes, however, when right-wing propagandist LibsOfTikTok tweets about them, resulting a deluge of (mostly hostile) replies and mentions. Here's what that looks like in graph form.
cc:
@ZellaQuixote
Misleading pics feature in the "cement milkshake" narrative.
@javier_pan01
's image of a supposedly milkshake-induced chemical burn is a stock photo, and
@JackPosobiec
's speculation that a specific picture shows folks mixing cement milkshakes has been debunked by the photographer.
There are multiple actual journalists reporting on this event, but for some reason everyone is sharing this tweet from an account made last month with an AI-generated face pic as its avatar.
Please be cautious about helping potentially inauthentic accounts build an audience.
FYI the "My Twitter Worth" thing that a lot of people are tweeting today is part of the Round Year Fun set of malware apps. DO NOT DO IT, and if you already have, revoke its access to your account so it doesn't take unwanted actions on your behalf.
If you already used one or more of the Year Round Fun apps, you can revoke its access to your account and prevent it from taking further actions on your behalf. (Screenshots show the steps for revoking access using the Twitter website, phone apps are similar.)
These six newly-created verified accounts have 977 followers in common. One is
@verified
(which follows all blue-check verified accounts). The other 976 were all created on June 19th or June 20th, 2021, and all follow the same 190 accounts.
#Astroturf
Multiple large right-wing accounts have tweeted about losing large numbers of followers over the last couple of hours. So far, it's unclear exactly what's going on.
Two of these six accounts (
@kayitlii
and
@aykacti
) have photographs of people as their profile pics. Despite the presence of the blue verification checkmark, neither image is likely to depict the account holder as both images appear to be stolen.
For the benefit of any and all wishing to better use Twitter's search feature for your own investigative purposes, here is a Twitter search cheat sheet.
Tip: if you want Twitter to be the "world's most accurate real-time information source" or indeed any kind of information source at all, you should probably actually allow people to view posts.
Twitter is on a mission to become the world’s most accurate real-time information source and a global town square for communication. That's not an empty promise.
That’s OUR reality.
@steph93065
@ZellaQuixote
We rounded up the other accounts frequently retweeted by the retweeters of
@steph93065
, then repeated the process on the retweeters of those accounts etc until we ended up with a relatively stable list of the top 100 most retweeted
#MAGA
accounts (based on this method, anyway.)
This botnet uses multiple varieties of GAN-generated profile pics. (GAN = "generative adversarial network", the AI technique used by .)
• 160 fake human faces (all female)
• 169 fake cats
• 183 anime pics
As always, be careful of "BREAKING" news that is unaccompanied by sources or evidence. No video or credible reporting has surfaced to support these claims of a Russian attack on Moldova, but big right-wing accounts are spreading it anyway.
These 976 accounts are part of an astroturf botnet consisting of (at least) 1212 accounts. The network is split into followers, which follow the aforementioned verified accounts as well as other members of the botnet, and followees, which are followed by the other bots.
Update:
@TwitterSafety
appears to no longer be removing death threats. The man being targeted here been forced to flee his home (). Choosing to leave these threats online is a sociopathic and evil decision.
This title is a lie. The reddit account you based it on does not belong to the shooter. This would be a good time to come clean with your readers and admit you publish the things you want to be true rather than bothering with that whole pesky fact-checking aspect of journalism.
Each of these four sites has an associated Twitter account:
@CostSmm
,
@smmgoal
,
@markenetspanel
, and
@socialboostx
. Unsurprisingly, they appear to have a lot of their own merchandise amongst their followers: thousands and thousands of accounts created in late 2022 with 0 likes.
It's 2023, and accounts with GAN-generated faces continue to receive "verification" checkmarks from Twitter Blue here and there.
#BlueCheckShenaniGANs
(These accounts were found by searching for common English words in combination with "filter:blue_verified.)
cc:
@ZellaQuixote
Here's a proposal for
@TwitterSafety
: modify the "Account suspended" screen to include a brief description of which rule/rules the banned account violated, and have "Learn more" link to the related section of the Twitter Rules.
(Images are mockups with made-up account names.)
Thus far, GAN-generated face pics (at least unmodified ones) always have the major facial features (particularly eyes) in the same position on each image. This is easy to see when one blends the images together:
If you "offered" and didn't take no for answer, that's not an "offer" - that's you deploying armed personnel in blatant defiance of the democratically elected local government.
“I offered
@DHSgov
support to help locally address the situation that’s going on in Portland, and their only response was: please pack up and go home. That’s just not going to happen on my watch.”
A rogue's gallery of dishonest influencers have spent the last couple hours attempting to smear former Twitter Trust and Safety head Yoel Roth as a pedophile via old out-of-context tweets.
Predictably, the far-right hate mob is now spamming Roth with death threats.
Fun fact: this pardon doesn't make you innocent. It makes you a propagandist for a wannabe autocrat who has tossed you a bone for his own self-serving reasons.
For the record, I have NEVER participated in any "online harassment campaigns" directed at opponents of Ron Desantis or anyone else.
I look forward to an immediate retraction from
@TheDailyBeast
of this false and defamatory reporting.
We've done a decent amount of research on the use of GAN-generated images over the last two years, mostly fake face "photos" such as those produced by thispersondoesnotexist(dot)com. Here are all of our related threads in one place.
cc:
@ZellaQuixote
Roughly a week ago, an AI for generating synthetic faces was made publicly available at thispersondoesnotexist(dot)com. We used it to generate 5000 sample faces. Pictured below is the result of blending all 5000 images.
cc:
@ZellaQuixote
Funny how so many of the folks who were up in arms about imaginary child abuse in the basement of a pizza parlor with no basement are completely silent when the government actually loses track of 1475 children.
Unsurprisingly, the same thing happens when LibsOfTikTok tweets about a school or school district - the school or district's previously obscure Twitter account gets swarmed by LibsOfTikTok's followers.
It's a Saturday in May, and a reply spam network is busy spamming people's replies with extremely similar tweets encouraging them to log into an obscure website with supplied login credentials. (The login info is the same in every reply.)
cc:
@ZellaQuixote
Unmodified GAN-generated faces (so far) have the telltale trait that the main facial features (especially the eyes) are in the same position on each image. This becomes evident when one blends the images together.
This spam network consists of (at least) 59645 Twitter accounts, mostly created between 2010 and 2016. All of their recent tweets were sent via the Twitter Web App. Some accounts have old unrelated tweets followed by a multi-year gap, which suggests they were hijacked/purchased.
It would appear that recently-banned COVID disinfo merchant Alex Berenson has a ban evasion account with a blue checkmark (
@GenRescue
, permanent ID 24014317).
@TwitterSafety
, you might wanna check this one out.
We'd been contemplating mapping out the backbone of the
#MAGA
Twitter network for a while, and upon seeing several tweets commenting on the speed with which
@steph93065
's replies went viral, we decided to start with that account and spiral outward.
cc:
@ZellaQuixote
Note that the tweet in question was posted via the "Twitter Web App" rather than one of the Twitter smartphone apps, meaning it was likely tweeted from a computer rather than the cell phone of someone on the scene. (TLDR, it's probably plagiarized from whoever took the video.)
29 of the remaining accounts have default profile pics and most of the others have pics that are used by multiple accounts in the botnet. We've seen networks that combine human, feline, and anime GAN pics with default and duplicate pics before:
This network uses three different types of GAN-generated profile pics:
• human faces (mostly female) - 2092 accounts
• cats - 2037 accounts
• anime characters - 1944 accounts
(GAN = "generative adversarial network", the AI technique used to produce the images)
Imagine pretending you care about the First Amendment until a judge rules that it means your favorite corrupt president can't block his critics on social media.
#AltWankers
Let’s pay tribute to
@BillDeplora
, a Navy veteran whose service to our country, at least based on appearances, would seem to include figuring out how to appear to be a propaganda bot.
Permanent IDs of the empty blue-check accounts followed by this botnet, in case they rename:
@aykacmis
1405267251662622733
@degismece
1405264617023098880
@anlamislar
1405264528024145921
@aykacti
1405263426461503489
@kayitlii
1405262586376040448
@donmedim
1405261641982300161
In one of the least surprising developments ever, 4chan trolls are creating fake left-wing Twitter accounts with GAN-generated faces as profile pics. Let's find some.
#ASeriesOfUn4ChanateEvents
cc:
@ZellaQuixote
The
@SherylLewellen
account (ID 1464674954272776200) has now removed the GAN-generated profile pic and changed its display name to "Reclaim America - PatriotFront(dot)US".
A whole lot of big accounts took the bait on this one.
#DontGetFooledAgain
Let's take a look at a new
#MAGA
account, created on December 24, 2018:
@emmaraya05
. In the space of four days they've gained over 2000 followers.
cc:
@ZellaQuixote
I've been meaning to do this for a while - brief thread on the basic tools I use. All of this stuff is free and open source, and all of it works on Windows, MacOS, and most variants of Linux.
You have this exactly backwards. Infowars is not a critic of state authoritarianism - it is a purveyor of shameless propaganda on behalf of the wannabe authoritarian who is currently the head of state in the US.
#FactsMatter
#StuckInAnEmbassy
Jones responds to Google, Youtube, Apple, Spotify censorship. Infowars pushes a libertarian right world view, with doses of conspiracy, but it is also frequently the single most influential populist critic of state authoritarianism in English, of which the following is typical.
Funny how this particular case has completely failed to show up on Mike Cernovich's timeline. You'd think it'd be right up his alley.
#FollowTheWhiteRabbit
@GregRubini
@ZellaQuixote
Tweet IDs encode several pieces of data, including a number identifying the datacenter that processed the tweet. Since distance affects network speed, it would be logical to handle tweets in datacenters on the same continent as the user. Can we find evidence that this is so?
In the wake of the attacks on Andy Ngo at yesterday's counterdemonstration against the Proud Boys in Portland, the claim that the thrown milkshakes contained quick-drying cement has circulated widely on right-wing Twitter.
cc:
@ZellaQuixote
@Jim_Jordan
Interesting. I just ran the following four searches:
from:Jim_Jordan
from:DevinNunes
from:RepMarkMeadows
from:RepMattGaetz
Spoiler alert: none of the accounts you're claiming are shadowbanned actually are (quality filter is on, too).
A new account named
@kamala_wins47
has racked up 56K followers in just three weeks, and is now claiming to be affiliated with the DNC.
In reality, this account appears to be nothing more than the latest version of a banned T-shirt spam account,
@emywinst
.
cc:
@ZellaQuixote
@Emywinst
Selling T-shirts appears to be the goal of
@Emywinst
's video plagiarism. Each post containing a stolen video has an accompanying self-reply containing an ad for a T-shirt and a link to a site where the shirt can be purchased.
Attorney Natalia Veselnitskaya of Trump Tower meeting fame has apparently developed a fan club in the form of Twitter accounts with no followers who love to sing her praises to a variety of celebrity accounts.
@BillBrowder
cc:
@ZellaQuixote
Very few of the accounts in this network have tweeted. The majority of the tweet content is spam in Korean sent via automation service dlvr(dot)it promoting a website. As always, be wary of clicking links to unknown sites posted by dodgy accounts.
Speaking as a researcher who has done a massive amount of unpaid work using these APIs over the last six years (and drawn attention to millions of inauthentic accounts), this is one of the most asinine and blatantly anti-transparency decisions you could possibly make.
Despite being at least seven years old and having few or no tweets, these accounts have somehow accumulated over 50,000 followers each. In an interesting twist, almost all of those followers were created in November 2022.
cc:
@ZellaQuixote