On the Left, there’s been a temptation to dismiss the revelations about Twitter’s internal censorship system that have emerged from the so-called Twitter Files project. But that would be a mistake: the news is important and the details are alarming.
The Twitter Files give us an unprecedented peek behind the curtain at the workings of Twitter’s opaque censorship regime. (Tayfun Coskun / Anadolu Agency via Getty Images)
The so-called Twitter Files, which started being released at the start of December, have so far generated a lot more discussion of the metacontroversies surrounding their release than of what’s actually in the “files” themselves: controversies about who released the files, who reported on them, the way they were reported, the wrong-headed political beliefs of some of those involved in the reporting. That’s too bad, because for all its very real faults, the Twitter Files story is an important and consequential piece of reporting that everyone — particularly on the Left — should be paying attention to.
Make no mistake: while some criticisms of the project coming from left of center certainly have merit, that doesn’t mean the disclosures aren’t important, or that the accuracy of the information contained in the files is somehow undermined by the political slant of some of those reporting on it. The Twitter Files give us an unprecedented peek behind the curtain at the workings of Twitter’s opaque censorship regime, and expose in greater detail the secret and ongoing merger of social media companies and the US national security state. And while Bari Weiss may not be interested in them, there are major implications for the Left.
The FBI and CIA Are Deeply Enmeshed
Arguably the most important revelations concern the creeping enmeshment of Twitter and other tech firms with intelligence and law enforcement, something we’ve gotten a glimpse at previously, but is laid bare in more detail here.
The files confirm something that we’d previously only learned through a lawsuit: that Twitter and other tech executives were having regular monthly and even weekly meetings with not just the FBI and its Foreign Influence Task Force (FITF) — the eighty-person counterintelligence division made to combat foreign disinformation campaigns — but pretty much every security agency under the sun. Besides the bureau, those named include the Department of Homeland Security (DHS), the Office of the Director of National Intelligence, state governments like that of California, the justice and state departments, even the National Security Agency (NSA), which requested to be included in a special Signal channel set up for the election period to let government agencies feed information to social media companies, or “the industry.”
One of those agencies happened to be OGA, or “Other Government Agency,” something both Twitter’s emails and former officers make clear was the CIA, which Matt Taibbi writes was “nearly always” present at regular FITF meetings with every major tech firm you can think of. Far from merely listening, the CIA was an active participant, regularly kicking off FITF meetings with a briefing, according to Taibbi, and sharing intelligence through the bureau and FITF with a list of firms as broad as including Wikimedia, “to develop potential investigative leads.” In one email, then head of trust and safety Yoel Roth made clear he regarded giving information to the FBI as giving information to the IC [“intelligence community”] by proxy.
This relationship went deep. Twitter employed so many ex-FBI personnel that, according to Michael Shellenberger — one of the conservative writers given access to the files — they had their own private Slack channel. They even created a “‘Bu to Twitter’ translation chart,” a cheat sheet for FBI alumni to translate former bureau jargon into its Twitter equivalent (“BLUF (bottom-line up front)” at the FBI was “TL;DR (too long, didn’t read)” at Twitter, for example).
In the run-up to the election, the FBI gave Twitter executives temporary security clearances and shared classified information with tech firms, stressing that there were “no impediments” to such sharing. A special platform named Teleporter was created to let the FBI send its reports to Twitter, similar to the special portal we learned the bureau uses to request throttling of content by Facebook. At one point, the FBI’s Elvis Chan even asked Twitter to provide “any location information associated” with a list of accounts singled out for spreading election misinformation.
The result was a deluge of censorship requests from the FBI. The bureau, Taibbi reports, sent over “lists of hundreds of problem accounts” to Twitter executives, often so lengthy they came in the form of Excel spreadsheets, and “thousands of mostly domestic reports,” despite the fact that the FITF was meant to be focused on foreign influence.
The significance of this should be obvious. Many shortsightedly dismissed the first Twitter Files release about Twitter’s outrageous censorship of the New York Post on the grounds that because the government wasn’t involved, such censorship shouldn’t concern us. That premise is wrong, but even if we were to accept it, the fact is that the FBI and a host of other shadowy security agencies are clearly knee-deep in the decisions over what Twitter and other companies decide to censor.
There are also questions about the role of agencies like the CIA and NSA. These are meant to be foreign-facing spy services that are least nominally barred from turning their powers on Americans, and yet here they are clearly taking part, to different degrees, in domestic-based operations.
Finally, for anyone with even a cursory knowledge of the distant and recent histories of the CIA and FBI — an agency that privately labels anti-police brutality protesters “black identity extremists” to be spied on and has a long track record of targeting young and often disadvantaged Muslim men with predatory entrapment schemes — the fact they’re playing this extensive a role in deciding what social media companies censor should be disturbing.
Twitter Admitted “Foreign Influence” Accounts Had Little Effect ― but the FBI Pressured Them to Do More
While this enmeshment should be alarming, what’s abundantly clear from the documents is that the phenomenon these agencies were meant to be combating — foreign disinformation — is not. In a routine that became familiar during the “war on terror,” when finding itself short on actual threats, the FBI desperately looked for and sometimes invented such threats, and pressured Twitter personnel to do the same.
The emails inadvertently suggest how marginal and ineffective the accounts that were genuine Russian disinformation efforts actually were. Describing Russian accounts they’d purged in advance of voting in 2020, one internal email reports some decidedly underwhelming figures: 345 had been inactive, and the ten that were active “had little reach and low follower counts.”
In fact, much of the supposedly suspicious activity had nothing to do with adversarial governments at all. The FBI repeatedly passed on suspicions and forwarded news articles alleging sinister foreign hands behind the George Floyd protests, the existence of pro-Trump African Americans, and various suspect accounts, only to be told they were all of domestic origin. One analyst “found no links to Russia,” they wrote, and neither did another worker who carried out a thorough double-check. “I can brainstorm with [redacted] and see if we can dig even deeper and try to find a stronger connection,” the analyst helpfully offers.
As that line suggests, the FBI wasn’t shy about making known its unhappiness at Twitter’s failure to find such foreign threats, nor about pressuring it to try harder. In July 2020, the FBI sent the company a lengthy questionnaire demanding it explain its conclusion that it “had not observed much recent activity from official propaganda actors on your platform,” and included a list of studies that are “in contrast to your own analysis.” A “perplexed” Roth commented the questions were “more like something we’d get from a congressional committee than the Bureau,” and declared he was not “comfortable” with the implications of “state-controlled media” they carried. In another email, an executive says he left alone an account called InfoBRICS — since the Russian government is openly a member of the economic organization associated with the account — but noted that “our window on that is closing, given that government partners are becoming more aggressive on attribution.”
The bureau likewise pressured Twitter to relax its privacy standards and hand over more user information. It asked executives if they would revise their terms of service to effectively allow intelligence agencies to scoop up open-source data more easily (they said no), and if they’d share information about accounts using VPNs, which are used to mask online activity (also no).
“We have seen a sustained (If uncoordinated) effort by the IC to push us to share more info & change our API policies,” public policy director Carlos Monje wrote in January 2020. “They are probing and pushing everywhere they can (including by whispering to congressional staff). We should stay connected and keep a solid front against these efforts.”
This was going on at least as late as this past August, when an executive cautioned the bureau was going to use an upcoming meeting to try “to convince us to produce on more FBI [emergency disclosure requests],” which permit the FBI to get data without a warrant. In a prep call, the FBI “repeatedly emphasized Twitter’s lower level of compliance in comparison with other platforms” and that they planned on bringing statistics to make their case.
When it wasn’t pressuring Twitter to find evidence of foreign threats, the FBI was setting its crosshairs on low-level satire for lack of anything else to do. Many of those flagged and even suspended for election misinformation were low-follower accounts making the same lame joke telling people to vote on the wrong day with next to no engagement, a trend that ensnared Democractic-aligned accounts the same as others. The bureau was “just doing keyword searches for [terms of service] violations,” one legal executive wrote, with even former FBI-general-counsel-turned-Twitter-attorney Jim Baker deeming it “odd that they are searching for violations of our policies.”
And it wasn’t just the FBI. “The Biden team was very not satisfied with Twitter’s enforcement approach as they wanted Twitter to do more to deplatform several accounts,” Twitter’s head of US public policy wrote this December. “Because of this dissatisfaction, we were asked to join several other calls. They were very angry in nature.” The White House in particular pressured Twitter to ban Alex Berenson, a former New York Times reporter who has spread a variety of dubious information about COVID and vaccines throughout the pandemic.
That Berenson espouses nonsense doesn’t make this revelation less disturbing: to accept this precedent is to accept that any future White House can press Twitter or any other social media firm to remove accounts it decides is spreading misinformation.
Twitter Censored US-Labeled Foreign Propaganda While Giving the Homegrown Version a Pass
There was a stark contrast between how the FBI treated accounts deemed to be foreign propaganda, or even just foreign state-directed, and how it treated their US equivalents, which were given a free pass.
Intelligence reports flagged dozens of YouTube videos and posts allegedly linked to a Russian troll farm that “highlighted predominantly anti-Ukraine narratives,” and listed more than one thousand accounts that it determined were “linked to the [Venezuelan president Nicolás] Maduro (VEN) & [Cuban president Miguel Mario] Díaz-Canel (CUB) regimes” and were “propagating anti-Bolsonaro/pro-Lula hashtags” (referring to Brazil’s outgoing far-right president, Jair Bolsonaro, and his left-wing challenger, Luiz Inácio Lula da Silva, in this year’s elections, respectively).
Meanwhile, there is evidence that Twitter treats US government-linked propaganda very differently. Drawing on the material in the Twitter Files, the Intercept’s Lee Fang found that the company exempted from censorship a number of accounts at Washington’s request that the Pentagon used to try shape public opinion in countries like Iraq and Syria, including government-created news outlets, and let them operate freely for years on the platform even after the Pentagon hid its connection to the accounts, just as the FBI accused Russia of doing.
In one instance, a US Central Command official emailed Twitter to ask directly to “whitelist” certain Arab-speaking accounts possibly flagged as bots that “we use to amplify certain messages” and which “had built a real following and we hope to salvage.” One of those accounts, no longer in service, promoted the US-backed Saudi-led war in Yemen, and painted US drone strikes as “accurate” and only killing terrorists instead of civilians.
In other words, the creeping merger of the national security state with Twitter doesn’t just bring up issues of political censorship. It also suggests that the website supposedly meant to be the “global public square” is being used as a geopolitical tool in the service of one government’s foreign policy interests.
Twitter Has All Manner of Tools to Quietly Throttle Users ― and Its Rules for Doing So Are Inconsistent and Ad Hoc
Finally, the tranche of documents shows that, even without government pressure, Twitter’s executives had a habit of leapfrogging their own rules and established procedures to carry out censorship of what they deemed to be misinformation.
Through “visibility filtering,” Twitter, it turns out, has a variety of tools it can use to “suppress what people see to different levels,” as one employee told Bari Weiss, without outright suspending or banning it. Individual account pages used by Weiss as examples show things like “trends blacklist,” “search blacklist,” and “do not amplify,” all aimed at making it harder for users to search or discover certain accounts.
Weiss has, predictably, focused on right-wing hobbyhorses to make her point, but given that the team responsible for this function “often handled two hundred ‘cases’ a day,” according to Weiss, it seems likely there are left-leaning examples that she and other conservative reporters didn’t care to search out.
As with the Hunter Biden laptop story, Twitter executives often seemed to be actively looking for justifications to carry out the censorship they wanted to do anyway. Roth told a colleague the company “has used technicality spam enforcements as a way to solve a problem created by Safety under-enforcing their policies.”
In the process, executives seem to have been regularly generating creative reinterpretations of existing policies, or simply making up new rules on the fly. For example, the decision to ban Donald Trump came just a day after the company created a new five-strike system for permanent suspensions — and after Roth acknowledged Trump had one more strike left. It led one employee to call the ban a “one-off ad hoc decision,” not the only example of Twitter employees describing acts of censorship as “one-off decisions.” To carry out Trump’s ban, Roth also made clear that “in this specific case, we’re changing our public interest approach for [Trump’s] account,” the company’s policy of public-interest exception, for which it specifically cites “tweets from elected and government officials.”
In several examples, Trump tweets initially flagged as problematic turned out to be, strictly speaking, factually accurate or acknowledged as not being rule violations, with Twitter’s head of legal, policy, and trust Vijaya Gadde suggesting instead to interpret the latter as “coded incitement to further violence.” Not having “a firm policy basis for action” on actor James Woods’s account after he tweeted the company was “suppressing” Trump’s account, they resolved to “hit him hard on future vio[lation] with firmer basis.”
By contrast, the company went easy on tweets from anti-Trump accounts claiming that Trump and recently confirmed Supreme Court justice Amy Coney Barrett would steal the election. In one case, Roth stepped in directly to have the company reverse a warning label placed on a tweet by former Obama attorney general Eric Holder saying that it was “too late to use the mails” to vote and urging people to vote in person, while charging the Supreme Court was trying to “deprive you of your most precious right.” (Incidentally, undermining “critical functions carried out by other key democratic institutions, such as the courts,” [emphasis added] was one of the types of dis- and misinformation the DHS specifically singled out as a priority to combat).
None of this absolves noxious right-wing voices or their many dubious claims about the 2020 election. Rather, it shows the inherently subjective nature of deciding what’s misinformation and what isn’t, and how this makes the process inevitably inconsistent and shaped by the biases of whoever’s enforcing it — enforcers who may not always hold liberal values or pro-Democratic political leanings. More than that, it shows that Twitter’s “content moderation” was often far from balanced and predictable — as a policy is meant to be — and instead frequently seemed to be decided on a whim.
No Cause for Comfort
Many have been quick to dismiss or ignore all of this because so many of the illustrative examples disclosed so far have been genuinely repugnant people or politically unsympathetic. But this is a classic mistake.
Political suppression, if it’s allowed to get a foothold, always starts at the outer edges of what’s acceptable before bit by bit expanding to envelop legitimate but dissident voices that it was supposedly never meant to target. This is why liberals and leftists fiercely denounced a panoply of “war on terror” overreaches that were at first aimed at actual terrorists — not because they agreed with terrorists or viewed their words and deeds as legitimate, but because it was understood such powers could easily be abused and weaponized against others.
(Taibbi noted in response to a question of mine on Callin that “Palestinian accounts are often the canary in the coal mine for new moderation or censorship techniques,” but he added that they weren’t featured in the first dataset he pulled. “I’m hoping that once we’re able to have a more global look at things that we can start looking at that,” he said.)
But whether or not Taibbi and others dig up reams of cases of Left-targeting censorship, many things could change in the years ahead that make the existing status quo more perilous for leftists: Twitter could start hiring conservatives; it could overcorrect in the face of right-wing criticism, or buckle under mounting FBI pressure; national security agencies could embark on a fiercer crackdown of the Left; a Republican could return to the White House; or a GOP Congress could follow Democrats’ example and pressure tech firms to censor what conservatives consider dangerous speech, to name a few possibilities. The best defense against this is to start opposing these trends now, not when it’s too late.Original post