Election Security

How Might the Sleeper Agents From “The Americans” Interfere in the Election?

By Herb Lin, Steven Weber
Tuesday, August 4, 2020, 2:55 PM

Setting the Stage

As the November 2020 presidential election approaches, it is worth imagining how a foreign adversary might attempt to intervene in the domestic political process. We have no evidence that any of the precise things we consider in this essay are actually happening—though some may well be. They are based on a review of what we know to be possible and plausible given what has occurred in the past and the vulnerabilities we can see clearly today. We don’t make specific assertions about the likelihood of any of these efforts or the probability of any having a meaningful effect.

Put simply, these things could happen; they might or might not work as intended, and, more importantly, we might or might not know of their existence immediately (or ever). Such uncertainty is, of course, part of the operation’s impact itself. We believe that what follows is a realistic, though ambitious, scenario for what could happen should the capabilities of Russian-style information warfare be deployed aggressively against the United States between now and January 20, 2021.

Our analysis is based on several assumptions. First, foreign actors will take advantage of a variety of technological instruments at their disposal, including social media, search engines, email and the like. Second, they will leverage the cacophony of political voices in the U. S. across the entire political spectrum. Third, they will prefer to exploit existing societal fault lines in the U. S. rather than attempting to create new ones.

It’s the third assumption that makes this more than just rampant speculation where the most imaginative and outrageous ideas get the most attention. We are assuming, instead, that adversaries are cost-sensitive and resource-constrained, and are looking for easy paths to disruption rather than the most sophisticated or elaborate ones. This, of course, is consistent with what we know about the cybersecurity landscape overall. Highly advanced attack modalities are the most scientifically interesting, but they are not generally the most criminally or politically interesting, simply because much less sophisticated and cheaper attacks will get the job done. The same is almost certainly true of attacks on the 2020 election.

The protagonist in this piece is named Philip Jennings, coincidentally the same name as the lead character in the FX television series The Americans. In that series, Phillip is a clandestine Russian KGB agent living in the U.S.—an “illegal” in spy novel parlance—and his cover is that of a suburban dad and travel agent in Washington, D.C.

Philip would likely have three goals in mind that would serve Russian interests:

  • To help lay a foundation for delegitimizing a Joe Biden presidency should Biden assume office on Jan. 20, 2021; and for the Republican Party to begin a de facto campaign to launch impeachment proceedings against Biden on Jan. 21, 2021—much as some of the Democratic establishment did with Donald Trump in 2017.
  • To increase the likelihood of a second presidential term for Donald Trump and to assist in Trump’s efforts to minimize accountability for the misdeeds and misjudgments of his various friends, supporters and political allies, while highlighting the same types of misdeeds for Biden’s associates and family.
  • To increase and amplify political and social polarization regardless of who wins the election in order to make Americans even more hostile and more angry toward each other, to increase cynicism about U.S. democracy at home and abroad and to enhance “epistemic fracture” that deprives political debates of a shared foundation of facts.

Activities that advance any of these goals could be modified in ways that enhance and multiply negative economic consequences, conferring an additional advantage to Russia since an America in recession is likely an America in continued retreat from global engagement regardless of who occupies the White House. Some of these activities would also detract from the effectiveness of a national response to the continuing threat and spread of COVID-19. This would likewise provide an advantage to Russia, helping to turn world attention away from the spread of the novel coronavirus in Russia, and focus it directly on America’s catastrophic response. Other activities would help to focus global attention on the grassroots anti-racism protests and demonstrations across the U.S. that do not appear to be stopping anytime soon.

Philip is not capable of doing a significant fraction of what we describe below by himself. He will do some of it. But more importantly, because he has been in the United States for a long time, he has an intimate and intuitive familiarity with many aspects of the American political and cultural scene. Additionally, he has had the opportunity to immerse himself in a variety of online communities under different identities in anticipation of just such activity when the time is right. (Such careful preparation—undertaken years in advance—gives Philip a major advantage over others who might wish to act more opportunistically.) With secure communications channels back to Russia, he can suggest a range of actions that might be taken and provide strategic advice and guidance to his Russian handlers, who are capable of mobilizing a large number of other agents to act in accordance with such advice.

The most scientifically interesting cyberattacks are those against electronic voting machines, voter registration databases and vote-reporting systems, but Phillip probably isn’t going to aim at those tougher targets—simply because he doesn’t have to. As a suburban dad who runs a travel agency as a cover, he’s got plenty of access to much lower-hanging fruit that can be targeted with a moderate mix of technology and social engineering available to almost anyone. It’s worth noting that because Philip is in fact an American citizen, many of the activities we’ve imagined him taking here are protected by the Constitution.

Delegitimizing Joe Biden as President

As of early August, polling data shows Biden with a substantial lead over Trump, both nationally and in a state-by-state match-up. Although polling data before Labor Day is generally discounted as a predictor of election outcomes (and in 2016 Trump was substantially behind at this time yet went on to win the election), Philip would be foolish to discount the possibility that Biden could indeed be elected president in November. It is thus in his interest to plan for such an outcome, because a president deemed illegitimate by a significant fraction of the country will be severely constrained in the actions he could take to counter adversarial Russian activities. Here are some activities that Philip could orchestrate to delegitimize a Biden presidency:

  • Sow distrust about the election, calling it rigged, stolen, manipulated or simply fraudulent. Elements could include making non-specific accusations or spreading rumors about voter fraud, tampering with electronic voting machines or voter registration systems and casting doubt about whether provisional ballots will be counted. Given existing doubts about voter fraud, particularly on the right, documented vulnerabilities in electronic voting systems and evidence of Russian penetration of voter registrations, and the relative novelty of provisional ballots, an information operation to amplify these issues is likely to find receptive audiences. Amplification need not rely on a substantial volume of evidence. Even if only a few isolated instances actually happen—as they do in all elections—endless repetition of those stories can do much to keep them in the public eye of selected audiences. Entirely false stories can also be used and publicized, and, if and when taken down, new false stories with similar themes could be generated and used. Phillip knows that a looming question mark about the possibility of fraud is equally or more powerful than a known and certain manipulation, and he knows that efforts to suppress stories of this kind have been shown in some cases to actually multiply their presence on social media.
  • Challenge the eligibility credentials of Biden’s running mate, depending on who it is. For example, Senator Tammy Duckworth—whom some have suggested as a plausible vice-presidential choice for Biden—was born in Thailand but with a father who was an American citizen. Under U.S. law, a child born under such circumstances is an American citizen from birth, but whether such a person satisfies the Constitutional requirement that the president be a “natural born” citizen of the United States is a contested point with few legal precedents and little legislative history to provide guidance. Of course, faked documents or testimony could be produced to cast doubt even on individuals that unambiguously satisfy the constitutional requirement. President Trump’s “birther” campaign against Barack Obama provides a ready blueprint for how to do this.
  • Publicize and comment on “old” Biden stories involving insensitivity towards women, financial improprieties or scandals involving corporate influence. Some of these stories may have a grain of truth to them, while others would be invented out of whole cloth. But as long as repeating them and calling attention to them through social media and targeted advertising is easy to do, neither their truth value nor their age is as relevant as their saliency to politically-attuned audiences. These stories would lay the political groundwork for a future impeachment that begins on day one of the Biden term, mirroring what some congressional Democrats said about Donald Trump in 2017, or at least for the start of investigations that may lead in that direction. Phillip might launch a “president-for-now” meme through some simple and creative TikTok videos, which of course would also divert attention from Russian election interference toward U.S.-China technology decoupling. Again, the #NotMyPresident story from 2017 provides the blueprint.
  • Discredit and/or disrupt ongoing prosecutions and investigations of Trump allies and Trump himself, arguing that they are politically motivated, without foundation and retaliatory. President Biden will likely want to draw a bright line between the past four years and his new administration; Philip wants to blur that line as much as possible and keep the cleavages of the Trump era as live as he can. He can do this by drawing continued attention to minor missteps and errors in the investigations of Trump and his associates and inflate their significance through social media. He can attack email systems of political opponents (for example, by finding and releasing selected emails that might incriminate them) and try to persuade career officials or lower-level career civil servants to modify or destroy evidence; even the hint of evidence manipulation is helpful. A bolder approach would be to flood investigators with fabricated evidence contradicting prosecutorial and investigatory conclusions using relatively simple deepfake technologies. For example, given access to an email account, it is also possible to insert forged emails that were never actually written by the authorized users of the account.
  • Conduct an information campaign to advocate for blanket pardons of Trump associates and even Trump himself for any prior activities that may have violated federal law. Such a campaign would highlight the perceived illegitimacy of investigations now underway and the expectation of unfair treatment in a Biden administration.
  • Blame the inevitable embarrassing and politically damaging incidents in the Trump campaign or the Republican National Committee on Democratic-leaning hacktivists. The credibility of this tactic could be enhanced by conducting false-flag operations in which Philip would recruit individuals to volunteer for these campaigns and engage in some minimally damaging activities; subsequently revealing these activities would provide some truth to fuel suspicions that other subsequent embarrassing incidents were also fabricated by opponents. This has the dual effect of giving the Trump campaign a convenient “free pass” when the incumbent goes off-script; and it provides an opportunity for the Republicans to deploy the “X-gate” script as a plausible accusation of Democratic Party meddling along the way.
  • Target right-wing sympathizers in the U.S. Armed Forces (especially the Army and Marine Corps, both active-duty and reserve/National Guard) to encourage them to support Trump both before and after the election, even if Biden wins. Until January 20, Trump is commander-in-chief, and all military personnel are duty-bound to obey his orders unless they are illegal. But whether they obey his orders enthusiastically or grudgingly makes a big difference, and Philip’s encouragement would be aimed at promoting enthusiastic support. Democrats and much of the press will be highly sensitized to even the smallest signals of military “activism” and a breakdown of civilian control during the transition period, as narratives about military intervention and even coups are already circulating and make for powerful stories that attract readers both in social media and traditional media. If Trump loses, it is in Philip’s interest to have the pre-election spectre of the armed forces backing his claim that Biden won the election fraudulently ready to go, even if the armed forces would likely not ultimately back his bid to stay in the White House. Phillip can count on President Trump to amplify this narrative during his last month(s) in office.

Increase the Chances of Trump Winning

Given that recent polls show support for Trump has been declining nationally, in swing state and among several electorally-important demographics that supported him in 2016, Trump’s principal path to victory currently appears to depend on generating the largest possible turnout of his base—the voters that will never abandon him even if he “shoots someone on 5th Avenue in New York City,” as he said in 2016. The most reliable way to generate that turnout is to make Trump voters fear that Trump will lose the election, even if polls shift to show him ahead in early November. It is important to do this in a way that can suppress turnout for Biden (or at a minimum, not boost it) at the same time.

Here are some activities that Philip could orchestrate to increase the chances of a Trump win:

  • Stoke fear in Trump or Trump-leaning voters of racial groups like African Americans, Muslims and non-white immigrants. One approach is to highlight demographic trends that demonstrate that the white Trump constituency is losing out as the United States moves closer to becoming a majority-minority nation, creating a sense of “now or never,” “the longer we wait the worse it gets,” and “this is our last chance to keep it from happening.” Philip would encourage a slowly-escalating rhetorical style in ads targeted towards such voters that would reach its peak shortly before the election at a level just short of openly promoting violent acts against these groups. Coded messages (“dog whistles”) targeting any of these groups would be within the scope of such an ad campaign, and coding would provide a measure of deniability. Philip can rely on mainstream media outlets to magnify the intensity of these messages in ways that embitter just about everyone.
  • Spotlight any and all violence—no matter how minimal—when it happens, especially violence that involves (as instigators, victims or, even better, ambiguous roles) any of the racial minorities mentioned in the previous paragraph. Respond rapidly to such events to amplify their coverage among Trump and Trump-leaning voters, using tweets, memes, Facebook groups and posts, news stories and opinion pieces. Increase the likelihood of violent confrontations when possible through online organizing—for example, by scheduling events for both sides at the same time and the same place under different aliases. Use agents provocateurs to instigate and inspire vandalism and to taunt police as a part of peaceful protests, to highlight the “law and order” challenge to which Trump portrays himself as the solution.
  • Insinuate that U.S. government agencies reporting on unemployment and economic growth are biased against Trump, and that the “official” numbers just reflect the “opinions” of a deep state bureaucracy committed to forcing Trump out of office precisely because he wants to break its power. Alternatively, or in addition, Phillip might seed the creation of an “independent” data organization providing a narrative arguing that U.S. government statistics are manipulated to make Trump look bad and supplying alternative statistics claimed to be derived from open data sources and machine learning inferences—and thus more modern and “scientific” than anything from the Department of Commerce. This organization would include anecdotal stories (some of which may even be true) about the brilliant people behind these alternative statistics and their commitment to disruptive innovation.
  • Raise public hopes about the prospects for effective vaccines and treatment, such as papers that have not yet cleared peer review and vaccine candidates that have not completed adequate Phase III testing (or for which Phase III data are not entirely definitive). Meanwhile, cast doubt on those following the scientific process, criticizing changing advice as the science advances (“flip-flops”) and complaining about the slowness of the process. Plant stories about clinical trial subjects—all of whom just happen to be Democrats—fabricating claims of vaccine side effects, insinuating that even the clinical trial recruitment process has been compromised and captured by Biden supporters. Philip’s goal here is to promote the narrative that Trump, as blustery and inconsistent as he may be as a public showman, is secretly competent behind the scenes, guiding Operation Warp Speed toward success—with a vaccine ready to come to market in less than a year.
  • Hack exit polling processes by flooding the Internet with a semi-random set of exit polling numbers that hint at surprises all over the place—for instance, high democratic turnout in Texas at the same time as Massachusetts reports surprisingly large numbers for Trump, thus creating the narrative that this is “an election like no other” in which patterns are being broken. A narrative of broken election patterns favors Trump as the self-proclaimed anti-establishment candidate running against Joe Biden, the ultimate establishment candidate. It thus lays some groundwork for later claims of fraud.
  • Intimidate COVID-19 contact tracers in ways that reduce knowledge of disease spread. Phillip could easily pump up the privacy rhetoric to reduce cooperation with COVID-19 contact tracers by planting stories, for example, about contact tracers targeting racial minorities, or ex-spouses, or competitive or unfriendly coworkers. Alternatively or in addition, he could encourage super-cooperation—telling those who test positive for COVID-19 to name anyone with whom they may have had even fleeting contact, thus artificially amplifying the number of contacts and wasting contact tracers’ time. As always, stories about invading children’s privacy will be highly effective.
  • Suppress votes for Biden in swing states. Philip’s efforts could be aimed at discouraging and/or intimidating likely Biden voters by casting doubt on the confidentiality of their votes coupled with personal threats aimed at “all those who voted for socialism and against America.” Trolls and bots could target registered Democrats with threats and intimidation—for example, they could conduct a social media campaign to suggest that anyone showing up to vote without proper ID or without being properly registered could be arrested on suspicion of fraud. On election day, Philip could encourage individuals to take advantage of open-carry laws and deploy in force carrying firearms to polling places to “stand guard.”
  • Suppress votes for Trump and disrupt election administration in states that are leaning heavily towards Trump. Because of this leaning, these suppression efforts will have no effect on the outcome in the state but will create and support a narrative that the Russians, or some other actor, are intervening in an even-handed way against both candidates.
  • Disrupt election administration in swing states by creating election-day confusion through the hacking of directions to polling places, faked changes in polling place addresses and suppression of poll workers at the polls through intimidation (e.g., threatening their families), harassment (e.g., slashing tires on cars belonging to poll workers), and false instructions (e.g., phoning or emailing them to tell them not to report). Because Philip knows that poll workers are often older people who are likely to be at greater risk from COVID-19 infection, a misinformation campaign that highlights the risk of catching the disease at polling places could contribute to a de facto labor shortage for managing election day. These activities would take some amount of advance preparation, but most of the information needed to conduct them is easily available either publicly or by hacking of those keeping track of poll workers.

Increase Polarization Around Wedge Issues

Philip would also have strong incentives to amplify polarization among Americans—even aside from the upcoming election—thus increasing the difficulty of communicating across partisan lines and of addressing hard problems. As this note is being written, three of the most significant wedge issues in American society are the pandemic itself, race relations and the economy.

On COVID-19, the wedge issue dividing Americans appears to be the tension between those who want to adopt the public health measures needed to reduce the spread of COVID-19 and those resisting the scope and scale of those measures. Phillip’s overarching goal here is to set up and exacerbate a fight between science and defiance—between the self-proclaimed defenders of virological and epidemiological knowledge, and the self-proclaimed defenders of American resilience who deny the power of a mere virus to take down the United States.

Philip is starting on fertile ground. As of July 2, 2020, 10.4% of Democrats compared to 81.2% of Republicans approved of the President’s handling of the COVID-19 outbreak. But the partisan divide is also reflected in other, more surprising ways. Polling in June and July 2020 showed that

  • 85 percent of Democratic-leaning adults viewed the coronavirus outbreak as a major threat to the health of the U.S. population, while 46 percent of Republican-leaning adults agreed.
  • 94 percent of Democrats self-reported they “always” or “very often” wore a mask when outside of their home, compared to just 46 percent of Republicans, while only two percent of Democrats self-reported “rarely” or “never” wearing a mask compared to 36 percent of Republicans.
  • 70 percent of Democrats vs. 34 percent of Republicans supported staying home except for essential errands.
  • 74 percent of Democrats vs. 33 percent of Republicans supported closing bars and restaurants.

Philip might thus suggest an information campaign to promote COVID-dismissive behavior among Trump supporters and a different one to draw attention to myriad examples of COVID-protective behavior among Democrats. He might try to shift the discourse from “Republicans don’t wear masks” to “minorities don’t wear masks.” He could encourage Trump supporters to exercise their civil rights by socializing normally—or even to do so excessively as explicit shows of defiance. As the availability of a reliable and effective vaccine draws closer, he would also stimulate anti-vaccine sentiments to increase vaccine hesitancy, especially among communities already suspicious of government vaccination programs.

As the pandemic gets worse, Philip would promote the Administration’s narrative that the response—including providing personal protective equipment to clinicians—is a state responsibility. At the same time, he would amplify the voices in states calling for a more robust federal response. Philip would also disparage public health experts (with Anthony Fauci as the primary target) who do not express great optimism about the evolution of the pandemic. Creating scandals (for example, circulating rumors about experts shorting equity markets before giving negative testimony) might also be useful.

Statistical models of disease spread leave Philip plenty of room to stoke defiance. Given the evolving state of scientific knowledge about the novel coronavirus and how it spreads, it would be easy to spin a defiant narrative. Perhaps the claim that there are a large number of asymptomatic infections is just nonsense—after all, we haven’t tested enough people to know. Or perhaps the notion of super-spreaders is just political—with the Democrats, afraid to gather for campaign rallies, propagating this myth merely to keep Republicans away from rallies that mobilize their base. And perhaps the data that shows most European states managing the pandemic so much more effectively is an indicator that U.S. data is (intentionally) flawed, not that the U.S. response has failed.

On race, it is clear that George Floyd’s death at the hands of the Minneapolis police has sparked a genuine grassroots movement that is strong, growing and unprecedented—at least as reflected in polling data suggesting that the substantial numbers of white Americans sympathize with sentiments expressed regarding the existence of systemic racism and police brutality against Black Americans. But it is also clear that a minority of the population—largely in Trump’s base—believe that the Floyd killing was an isolated incident committed by a few bad actors in police forces rather than a reflection of systemic issues, or even that the protests are unjustified because Floyd was a “violent criminal.” From Philip’s perspective, the lack of a middle ground or accepted compromise positions on what actions to take right now makes his job easier.

To exploit this issue, Philip would naturally seek to widen the chasm between these two sides. The simplest method for doing so would be to showcase to each side the (inevitable) excesses and extreme behavior of the other. In other words, Philip would publicize examples of the behavior of white supremacists to Black Lives Matter supporters and examples of violent behavior by Black Lives Matter supporters to Trump voters. In social and traditional media, he would highlight existing stories about excessive provocation, and then amplify the voices of those who make worst-case interpretations of behaviors and expressed opinions on both sides. His goal would be to create self-sustaining cycles of escalation between the two camps. The Harper’s letter of early July and the cycles of reciprocal outrage it provoked are an excellent template for this approach, as are the reported actions of federal law enforcement in Portland, Oregon, during mid-July. More rumors of unmarked “security forces” and “little green men” that make protestors “disappear” will be hard to prove false, and very effective.

On the economy, Philip could do many things to enhance existing cleavages. Fake stories are likely unnecessary: Philip can employ the very real and gut-wrenching tradeoffs that the pandemic has created in order to exacerbate resentment and anxiety. Philip could:

  • Plant stories about the “left behinds” and “the forgotten worker”—the people who lost jobs due to the pandemic that are not coming back, in the service and hospitality sectors, for example. Do everything possible to build on already-existing grievances here, such as by tying jobs lost by “real Americans” to immigrants, China, BLM supporters or racial minorities in the United States.
  • Highlight the increasingly obvious disconnect between capital markets (Wall Street) and the real economy (Main Street). Amplify the monetarists and others (like Nouriel Roubini, a.k.a. Dr. Doom) who warn of impending inflation and dollar collapse. Likewise, Philip could plant stories in newspapers and magazines about the collapse of the Weimar Republic.
  • Amplify voices calling for rapid reopening and a resumption of pre-COVID economic activity and also voices calling for caution and emphasizing the dangers of such resumption.
  • Do everything possible to further exacerbate the tech-lash dynamics that target Google, Facebook, Amazon, and their main shareholders—who can easily be portrayed as profiteering on the backs of the pandemic-induced recession.
  • Try to incite various “runs” on commodities—like the toilet paper and paper towel run that happened in March. Nothing creates economic and social anxiety quite like empty supermarket shelves.
  • Spread rumors about supply chain safety breakdowns in food—E. Coil in salad mixes and the like. Did workers infect the food supply in order to gain leverage with their employers for greater safety measures in production and processing facilities?
  • Flood Twitter, Instagram and other platforms with Depression-era bread line photos and the like, just to get those images in peoples’ heads. It would be ironic but effective for Philip to include images from the post-Soviet depression in Moscow, just to show how quickly an economy really can collapse.

Moving Forward

The purpose of this essay was not to put forward any new or profound solutions to forestall these kinds of activities from happening or to reduce their impact. Much of what we describe above depends not on the deliberate creation of false information but rather on focusing public attention on pre-existing narratives and stories that anger and polarize. As a result, recommendations that emphasize the exercise of critical thinking to determine whether an item of information is true or false are not particularly relevant—and certainly not in the time frame in which Phillip is operating.

From the standpoint of partisans on either side, it doesn’t matter much whether support or evidence for their point of view is authentic (e.g.., comes from genuine grassroots beliefs and activities) or inauthentic (e.g., is manufactured and fabricated in Moscow or somewhere else). So combatting the activities described above would require sincere desires on the part of partisans to repudiate inauthentic support, even assuming it could be identified (likely a bad assumption where Philip’s clandestine activities are concerned). There currently appears to be no prospect of any such repudiation. From Phillip’s perspective, the United States right now is indeed a very soft target.

We intended this essay to be a thought experiment that demonstrates how easy it would be to conduct such an intervention against the United States. Just a little bit of imagination, a little bit of technological savvy and a little bit of audacious ambition could together add up to an effective election interference program that would also further polarize America after the election regardless of the outcome.

What can be done to disrupt or degrade any of Philip’s activities? From our perspective, Philip’s most important leverage point comes from the amplification and force-multiplying effects of modern information technology, specifically social media. The social media firms have chosen to offer most of their products for public use at little or no cost, and political adversaries have learned to leverage the psychological vulnerabilities in humans who use these products to further their ends. Indeed, adversaries gain enormous advantage persuading and influencing unwitting parties to amplify their messaging for them.

If a physical product that is widespread in American society could be manipulated by an adversary—imagine an army of home-service robots whose operating systems could be attacked by a foreign power, and turned to hold families hostage inside their houses—it would be immediately addressed as a top-priority national security threat. But social media has long had a free pass for a number of reasons that apply to the information technology industry as a whole. Today, it is protected in distinctive and persistent ways because of its “speech” functions and the constitutional protections that these functions carry.

It thus stands to reason that the actors with sufficient reach and power to begin to clean up the information environment are in fact the social media companies themselves. What they lack are the economic incentives to do so. This is a business model problem first and foremost: the firms sell more ads (ads of all kinds, not just political ads) when users use their products more, so they have clear incentives to increase the duration and intensity of user engagement. And they know definitively from their data what has long been intuitively obvious: users find controversial, polarizing, divisive content more engaging and more emotionally compelling than measured, reasoned, civil dialogue.

Platform companies often assert that they could moderate the content they carry, in accordance with clear societal consensus on the nature of that content. They have in a number of instances demonstrated the capability to do so. It’s a hard technical problem, particularly given the extraordinary scale at which companies like Facebook and YouTube operate, but not an entirely impossible one. And the threshold for success doesn’t have to be 100% in order to make a meaningful difference. But in light of the economic incentives the firms have for carrying and promoting controversial content, they are in practice counting on the inability of a polarized and intellectually-fractured society to come to any kind of consensus regarding what content should be moderated.

Normal public relations-type communications strategies that the firms deploy to defend this position are increasingly implausible. In early July 2020, CEO Mark Zuckerberg said that Facebook would not change its content policies because of a threat to revenue, and a Facebook spokesperson followed up by claiming that “we make policy changes based on principles, not revenue pressures.”

Taking such statements at face value would suggest that it is the policy of the largest and most prominent social media company not to remove divisive, false or misleading content even if it contributes to a risk of immediate harm to American democracy because they default to a broad definition of the principle of free speech absent a firm societal consensus on a more precise definition. This may have been a good talking point for some time, but more careful reflection shows it to be cynically disingenuous and civically irresponsible. The U.S. Constitution protects speech, not any particular business model that deploys speech. What remains is for the companies to develop the political courage to be willing to try to mitigate such harms. Their attempts at mitigation will not get it right on the first try. But most of Silicon Valley was created on a “fail fast and iterate” mindset, in which first guesses are refined or even abandoned and thereby improved rapidly.

The last question we pose goes back to our protagonist’s handlers in Moscow: is now the right time to activate Philip? Philip’s handlers might very well see him as the human equivalent of a weapon that can really only be used once—effectively a zero-day exploit in the jargon of cybersecurity. It took extensive time and resources to put him in place, and it’s possible that at this very high level of activity his cover would be blown—after the election. The strategic decision his handlers have to make is whether the summer and fall of 2020 is a good use of this likely one-time resource. This thought experiment has led us to the conclusion that the answer to that question could very well be yes.