Social Media

Under Government Pressure, Twitter Suppressed Truthful Speech About COVID-19

The company's broad definition of "misleading information" and its deference to authority invited censorship by proxy.


Twitter's ban on "COVID-19 misinformation," which Elon Musk rescinded after taking over the platform in late October, mirrored the Biden administration's broad definition of that category in two important respects: It disfavored perspectives that dissented from official advice, and it encompassed not just demonstrably false statements but also speech that was deemed "misleading" even when it was arguably or verifiably true. In a recent Free Press article, science writer David Zweig shows what that meant in practice, citing several striking examples of government-encouraged speech suppression gleaned from the internal communications that Musk has been disclosing to handpicked journalists.

Twitter's moderation of pandemic-related content was intertwined with government policy from the beginning. Even before Joe Biden was elected president and his administration began publicly and privately demanding that social media companies suppress speech it viewed as a threat to public health, the company's guidelines deferred to the positions taken by government agencies such as the Centers for Disease Control and Prevention (CDC). And those rules explicitly covered "misleading information" as well as "demonstrably false" statements.

"We have broadened our definition of harm to address content that goes directly against guidance from authoritative sources of global and local public health information," Twitter said in a March 27, 2020, update. "Rather than reports, we are enforcing this in close coordination with trusted partners, including public health authorities and governments, and continue to use and consult with information from those sources when reviewing content."

That July, Twitter sought to clarify "our rules against potentially misleading information about COVID-19″ (emphasis added). "For a Tweet to qualify as a misleading claim," the company said, "it must be an assertion of fact (not an opinion), expressed definitively, and intended to influence others' behavior." Possible topics included "the origin, nature, and characteristics of the virus"; "preventative measures, treatments/cures, and other precautions"; "the prevalence of viral spread, or the current state of the crisis"; and "official health advisories, restrictions, regulations, and public-service announcements."

That was a very wide net, potentially encompassing anyone who questioned the CDC's ever-shifting guidance or criticized government policies, such as lockdowns and mask mandates, aimed at reducing virus transmission. While the intent requirement ostensibly allowed dissent as long as it was not aimed at influencing behavior, that limitation did not mean much in practice, since moderators were apt to infer the requisite intent when they encountered tweets that implicitly or explicitly deviated from the recommendations of "public health authorities and governments."

The "assertion of fact" requirement likewise proved malleable, since it applied even to true statements that were viewed as undermining compliance with those recommendations. And contrary to what Twitter's official rules said, even expressions of opinion could prompt action against tweets or users.

Consider a March 15, 2021, tweet in which the epidemiologist Martin Kulldorff responded to the question of whether "younger age groups" or people who had already been infected by COVID-19 "need to be vaccinated." Kulldorff's response: "No. Thinking that everyone should be vaccinated is as scientifically flawed as thinking that nobody should. COVID vaccines are important for older high-risk people, and their care-takers. Those with prior natural infection do not need it. Nor children."

Zweig reports that "internal emails show an 'intent to action' by a Twitter moderator, saying Kulldorff's tweet violated the company's Covid-19 misinformation policy" and claiming "he shared 'false information.'" But as Zweig notes, "Kulldorff's statement was an expert's opinion—one that happened to be in line with vaccine policies in numerous other countries."

Kulldorff's tweet nevertheless "was deemed 'false information' by Twitter moderators merely because it differed from CDC guidelines," Zweig writes. "After Twitter took action, Kulldorff's tweet was slapped with a 'misleading' label and all replies and likes were shut off, throttling the tweet's ability to be seen and shared by others, a core function of the platform."

Zweig says he found "numerous instances of tweets about vaccines and pandemic policies labeled as 'misleading' or taken down entirely, sometimes triggering account suspensions, simply because they veered from CDC guidance or differed from establishment views." Those actions were consistent with the Biden administration's understanding of "misinformation," which it defines as speech that deviates from a government-endorsed "scientific consensus."

Another example that Zweig cites: Last August, @KelleyKga, a self-described "public health fact checker," responded to another Twitter user's claim that "COVID has been the leading cause of death from disease in children" since December 2021. "What an excellent example of cherry picking!" @KelleyKga wrote. "If you narrow it down to only the specific months you specify, which include the largest Covid wave (seen across the world), AND you ignore all non-disease deaths, AND you ignore cancer, heart disease, SIDS, then COVID is 'leading.'"

That fact check included an image of the CDC's own data on causes of death among minors from December 2021 through the first half of 2022. Yet "internal records showed that a bot had flagged the tweet, and that it received many 'tattles' (what the system amusingly called reports from users)," Zweig says. "That triggered a manual review by a human who—despite the tweet showing actual CDC data—nevertheless labeled it 'misleading.' Tellingly, the tweet by @KelleyKga that was labeled 'misleading' was a reply to a tweet that contained actual misinformation."

Other examples that Zweig mentions include references to peer-reviewed studies of correlations between COVID-19 vaccinations and outcomes such as myocarditis, cardiac arrests, and temporary drops in sperm concentration and motility. Obviously, one can argue about the meaning and practical significance of those findings. But as Twitter saw it, merely mentioning them was "misleading," based on the inference that the authors of those tweets were trying to discourage vaccination.

Zweig notes that Twitter permanently suspended Rhode Island physician Andrew Bostom for several allegedly misleading posts, including his link to a study of sperm donors who had received mRNA vaccines and his citation of local and nationwide data indicating that "influenza is more lethal than covid-19 in children." After Bostom challenged his suspension, a Twitter review determined that three of the four tweets were not "misleading" after all.

An October 5, 2020, tweet by then-President Donald Trump further illustrates the potential breadth of "misinformation" as understood by Twitter employees. "I will be leaving the great Walter Reed Medical Center today at 6:30 p.m.," Trump wrote as he was recovering from COVID-19. "Feeling really good! Don't be afraid of Covid. Don't let it dominate your life. We have developed, under the Trump Administration, some really great drugs & knowledge. I feel better than I did 20 years ago!"

In an internal email that Zweig obtained, Jim Baker, a former FBI general counsel who at the time was Twitter's deputy general counsel, suggested that Trump's tweet was "a violation of our COVID-19 policy (especially the 'Don't be afraid of Covid' statement)." Yoel Roth, then head of Trust & Safety at Twitter, replied that Trump's "broad, optimistic statement" did not "incite people to do something harmful" or "recommend against taking precautions or following mask directives (or other guidelines)."

That exchange, which Zweig calls "surreal," suggests how subjective the application of Twitter's "misinformation" ban could be. As Roth saw it, Trump's tweet did not "fall within the published scope of our policies." But Baker apparently reasoned that people should be "afraid of Covid," the better to encourage compliance with government guidelines and mandates. He therefore surmised that Trump's message was "intended to influence others' behavior." And while Trump's tweet was an expression of "opinion" rather than an "assertion of fact," that distinction did not save Kulldorff.

Twitter's reflexive deference to government authority and broad definition of "misinformation" combined to invite the sort of meddling documented in the company's internal communications. When the Biden administration pressured Twitter and other platforms to crack down more aggressively on "misinformation," it argued that it was merely asking those companies to enforce their own policies. But given the power that the federal government has to make life difficult for Twitter et al. through castigation, regulation, litigation, and legislation, the administration's requests were tantamount to commands.

Federal officials expected meek obeisance, and that is typically what they got. Emails revealed during discovery in a First Amendment lawsuit filed by Louisiana Attorney General Jeff Landry and Missouri Attorney General Eric Schmitt show how keen executives at social media companies were to placate federal officials, especially after Biden accused Facebook of "killing people" by allowing the spread of vaccine misinformation.

Biden lodged that charge on July 16, 2021, the day after Surgeon General Vivek Murthy issued an advisory in which he urged a "whole-of-society" effort, possibly including "legal and regulatory measures," to combat the "urgent threat to public health" posed by "health misinformation." An executive with Facebook's parent company was eager to assuage the president's anger.

"Reaching out after what has transpired over the past few days following the publication of the misinformation advisory, and culminating today in the President's remarks about us," the executive wrote. "I know our teams met today to better understand the scope of what the White House expects from us on misinformation going forward."

At that point, Twitter had already gotten with the program. In April 2021, Deputy Assistant to the President Rob Flaherty sent colleagues an email about a "Twitter VaccineMisinfo Briefing" on Zoom. Flaherty said Twitter would inform "White House staff" about "the tangible effects seen from recent policy changes, what interventions are currently being implemented in addition to previous policy changes, and ways the White House (and our COVID experts) can partner in product work."

Twitter was eager to "partner" with the White House because the company had been repeatedly harangued for not doing enough to suppress "misinformation." Zweig quotes an email that Lauren Culbertson, Twitter's head of U.S. public policy, sent her colleagues in December 2022. It describes 2021 teleconferences between White House and Twitter officials. "The Biden team was not satisfied with Twitter's enforcement approach as they wanted Twitter to do more and to de-platform several accounts," Culbertson wrote. "Because of this dissatisfaction, we were asked to join several other calls. They were very angry in nature."

While "Twitter executives did not fully capitulate to the Biden team's wishes," Zweig says, the pressure had an impact on the platform's moderation decisions, leading not only to the banishment of anti-vaccine writers like Alex Berenson but also to the suppression of truthful statements by scientists and physicians. "As a result," Zweig notes, "legitimate findings and questions about our Covid policies and their consequences went missing."

The evidence collected by Zweig reinforces the point that the federal government has managed to impose censorship by proxy on social media platforms. Those efforts go beyond the subject of COVID-19, extending to election-related "misinformation," "hate speech," and possibly other categories of commentary that offend politicians or bureaucrats. Acting publicly and behind the scenes, "very angry" officials have indirectly limited speech that is indisputably protected by the First Amendment.