“ ‘Black Lives Matter’ doesn’t mean other lives don’t,” he wrote. “We’ve never had rules around what people can write on our walls,” the memo went on. But “crossing out something means silencing speech, or that one person’s speech is more important than another’s.” The defacement, he said, was being investigated.

All around the country at about this time, debates about race and politics were becoming increasingly raw. Donald Trump had just won the South Carolina primary, lashed out at the Pope over immigration, and earned the enthusiastic support of David Duke. Hillary Clinton had just defeated Bernie Sanders in Nevada, only to have an activist from Black Lives Matter interrupt a speech of hers to protest racially charged statements she’d made two decades before. And on Facebook, a popular group called Blacktivist was gaining traction by blasting out messages like “American economy and power were built on forced migration and torture.”

So when Zuckerberg’s admonition circulated, a young contract employee named Benjamin Fearnow decided it might be newsworthy. He took a screenshot on his personal laptop and sent the image to a friend named Michael Nuñez, who worked at the tech-news site Gizmodo. Nuñez promptly published a brief story about Zuckerberg’s memo.

A week later, Fearnow came across something else he thought Nuñez might like to publish. In another internal communication, Facebook had invited its employees to submit potential questions to ask Zuckerberg at an all-hands meeting. One of the most up-voted questions that week was “What responsibility does Facebook have to help prevent President Trump in 2017?” Fearnow took another screenshot, this time with his phone.

Fearnow, a recent graduate of the Columbia Journalism School, worked in Facebook’s New York office on something called Trending Topics, a feed of popular news subjects that popped up when people opened Facebook. The feed was generated by an algorithm but moderated by a team of about 25 people with backgrounds in journalism. If the word “Trump” was trending, as it often was, they used their news judgment to identify which bit of news about the candidate was most important. If The Onion or a hoax site published a spoof that went viral, they had to keep that out. If something like a mass shooting happened, and Facebook’s algorithm was slow to pick up on it, they would inject a story about it into the feed.

Facebook prides itself on being a place where people love to work. But Fearnow and his team weren’t the happiest lot. They were contract employees hired through a company called BCforward, and every day was full of little reminders that they weren’t really part of Facebook. Plus, the young journalists knew their jobs were doomed from the start. Tech companies, for the most part, prefer to have as little as possible done by humans—because, it’s often said, they don’t scale. You can’t hire a billion of them, and they prove meddlesome in ways that algorithms don’t. They need bathroom breaks and health insurance, and the most annoying of them sometimes talk to the press. Eventually, everyone assumed, Facebook’s algorithms would be good enough to run the whole project, and the people on Fearnow’s team—who served partly to train those algorithms—would be expendable.

The day after Fearnow took that second screenshot was a Friday. When he woke up after sleeping in, he noticed that he had about 30 meeting notifications from Facebook on his phone. When he replied to say it was his day off, he recalls, he was nonetheless asked to be available in 10 minutes. Soon he was on a video­conference with three Facebook employees, including Sonya Ahuja, the company’s head of investigations. According to his recounting of the meeting, she asked him if he had been in touch with Nuñez. He denied that he had been. Then she told him that she had their messages on Gchat, which Fearnow had assumed weren’t accessible to Facebook. He was fired. “Please shut your laptop and don’t reopen it,” she instructed him.

That same day, Ahuja had another conversation with a second employee at Trending Topics named Ryan Villarreal. Several years before, he and Fearnow had shared an apartment with Nuñez. Villarreal said he hadn’t taken any screenshots, and he certainly hadn’t leaked them. But he had clicked “like” on the story about Black Lives Matter, and he was friends with Nuñez on Facebook. “Do you think leaks are bad?” Ahuja demanded to know, according to Villarreal. He was fired too. The last he heard from his employer was in a letter from BCforward. The company had given him $15 to cover expenses, and it wanted the money back.

The firing of Fearnow and Villarreal set the Trending Topics team on edge—and Nuñez kept digging for dirt. He soon published a story about the internal poll showing Facebookers’ interest in fending off Trump. Then, in early May, he published an article based on conversations with yet a third former Trending Topics employee, under the blaring headline “Former Facebook Workers: We Routinely Suppressed Conservative News.” The piece suggested that Facebook’s Trending team worked like a Fox News fever dream, with a bunch of biased curators “injecting” liberal stories and “blacklisting” conservative ones. Within a few hours the piece popped onto half a dozen highly trafficked tech and politics websites, including Drudge Report and Breitbart News.

The post went viral, but the ensuing battle over Trending Topics did more than just dominate a few news cycles. In ways that are only fully visible now, it set the stage for the most tumultuous two years of Facebook’s existence—triggering a chain of events that would distract and confuse the company while larger disasters began to engulf it.

This is the story of those two years, as they played out inside and around the company. WIRED spoke with 51 current or former Facebook employees for this article, many of whom did not want their names used, for reasons anyone familiar with the story of Fearnow and Villarreal would surely understand. (One current employee asked that a WIRED reporter turn off his phone so the company would have a harder time tracking whether it had been near the phones of anyone from Facebook.)

The stories varied, but most people told the same basic tale: of a company, and a CEO, whose techno-optimism has been crushed as they’ve learned the myriad ways their platform can be used for ill. Of an election that shocked Facebook, even as its fallout put the company under siege. Of a series of external threats, defensive internal calculations, and false starts that delayed Facebook’s reckoning with its impact on global affairs and its users’ minds. And—in the tale’s final chapters—of the company’s earnest attempt to redeem itself.

In that saga, Fearnow plays one of those obscure but crucial roles that history occasionally hands out. He’s the Franz Ferdinand of Facebook—or maybe he’s more like the archduke’s hapless young assassin. Either way, in the rolling disaster that has enveloped Facebook since early 2016, Fearnow’s leaks probably ought to go down as the screenshots heard round the world.

II

The furious energy of this big bang emanated, in large part, from a brilliant and simple insight. Humans are social animals. But the internet is a cesspool. That scares people away from identifying themselves and putting personal details online. Solve that problem—make people feel safe to post—and they will share obsessively. Make the resulting database of privately shared information and personal connections available to advertisers, and that platform will become one of the most important media technologies of the early 21st century.

But as powerful as that original insight was, Facebook’s expansion has also been driven by sheer brawn. Zuckerberg has been a determined, even ruthless, steward of the company’s manifest destiny, with an uncanny knack for placing the right bets. In the company’s early days, “move fast and break things” wasn’t just a piece of advice to his developers; it was a philosophy that served to resolve countless delicate trade-offs—many of them involving user privacy—in ways that best favored the platform’s growth. And when it comes to competitors, Zuckerberg has been relentless in either acquiring or sinking any challengers that seem to have the wind at their backs.

In fact, it was in besting just such a rival that Facebook came to dominate how we discover and consume news. Back in 2012, the most exciting social network for distributing news online wasn’t Facebook, it was Twitter. The latter’s 140-character posts accelerated the speed at which news could spread, allowing its influence in the news industry to grow much faster than Facebook’s. “Twitter was this massive, massive threat,” says a former Facebook executive heavily involved in the decisionmaking at the time.

So Zuckerberg pursued a strategy he has often deployed against competitors he cannot buy: He copied, then crushed. He adjusted Facebook’s News Feed to fully incorporate news (despite its name, the feed was originally tilted toward personal news) and adjusted the product so that it showed author bylines and headlines. Then Facebook’s emissaries fanned out to talk with journalists and explain how to best reach readers through the platform. By the end of 2013, Facebook had doubled its share of traffic to news sites and had started to push Twitter into a decline. By the middle of 2015, it had surpassed Google as the leader in referring readers to publisher sites and was now referring 13 times as many readers to news publishers as Twitter. That year, Facebook launched Instant Articles, offering publishers the chance to publish directly on the platform. Posts would load faster and look sharper if they agreed, but the publishers would give up an element of control over the content. The publishing industry, which had been reeling for years, largely assented. Facebook now effectively owned the news. “If you could reproduce Twitter inside of Facebook, why would you go to Twitter?” says the former executive. “What they are doing to Snapchat now, they did to Twitter back then.”

It appears that Facebook did not, however, carefully think through the implications of becoming the dominant force in the news industry. Everyone in management cared about quality and accuracy, and they had set up rules, for example, to eliminate pornography and protect copyright. But Facebook hired few journalists and spent little time discussing the big questions that bedevil the media industry. What is fair? What is a fact? How do you signal the difference between news, analysis, satire, and opinion? Facebook has long seemed to think it has immunity from those debates because it is just a technology company—one that has built a “platform for all ideas.”

This notion that Facebook is an open, neutral platform is almost like a religious tenet inside the company. When new recruits come in, they are treated to an orientation lecture by Chris Cox, the company’s chief product officer, who tells them Facebook is an entirely new communications platform for the 21st century, as the telephone was for the 20th. But if anyone inside Facebook is unconvinced by religion, there is also Section 230 of the 1996 Communications Decency Act to recommend the idea. This is the section of US law that shelters internet intermediaries from liability for the content their users post. If Facebook were to start creating or editing content on its platform, it would risk losing that immunity—and it’s hard to imagine how Facebook could exist if it were liable for the many billion pieces of content a day that users post on its site.

And so, because of the company’s self-image, as well as its fear of regulation, Facebook tried never to favor one kind of news content over another. But neutrality is a choice in itself. For instance, Facebook decided to present every piece of content that appeared on News Feed—whether it was your dog pictures or a news story—in roughly the same way. This meant that all news stories looked roughly the same as each other, too, whether they were investigations in The Washington Post, gossip in the New York Post, or flat-out lies in the Denver Guardian, an entirely bogus newspaper. Facebook argued that this democratized information. You saw what your friends wanted you to see, not what some editor in a Times Square tower chose. But it’s hard to argue that this wasn’t an editorial decision. It may be one of the biggest ever made.

In any case, Facebook’s move into news set off yet another explosion of ways that people could connect. Now Facebook was the place where publications could connect with their readers—and also where Macedonian teenagers could connect with voters in America, and operatives in Saint Petersburg could connect with audiences of their own choosing in a way that no one at the company had ever seen before.

III

But McNamee didn’t say anything to anyone at Facebook—at least not yet. And the company itself was not picking up on any such worrying signals, save for one blip on its radar: In early 2016, its security team noticed an uptick in Russian actors attempting to steal the credentials of journalists and public figures. Facebook reported this to the FBI. But the company says it never heard back from the government, and that was that.

Instead, Facebook spent the spring of 2016 very busily fending off accusations that it might influence the elections in a completely different way. When Gizmodo published its story about political bias on the Trending Topics team in May, the ­article went off like a bomb in Menlo Park. It quickly reached millions of readers and, in a delicious irony, appeared in the Trending Topics module itself. But the bad press wasn’t what really rattled Facebook—it was the letter from John Thune, a Republican US senator from South Dakota, that followed the story’s publication. Thune chairs the Senate Commerce Committee, which in turn oversees the Federal Trade Commission, an agency that has been especially active in investigating Facebook. The senator wanted Facebook’s answers to the allegations of bias, and he wanted them promptly.

The Thune letter put Facebook on high alert. The company promptly dispatched senior Washington staffers to meet with Thune’s team. Then it sent him a 12-page single-spaced letter explaining that it had conducted a thorough review of Trending Topics and determined that the allegations in the Gizmodo story were largely false.

Facebook decided, too, that it had to extend an olive branch to the entire American right wing, much of which was raging about the company’s supposed perfidy. And so, just over a week after the story ran, Facebook scrambled to invite a group of 17 prominent Republicans out to Menlo Park. The list included television hosts, radio stars, think tankers, and an adviser to the Trump campaign. The point was partly to get feedback. But more than that, the company wanted to make a show of apologizing for its sins, lifting up the back of its shirt, and asking for the lash.

According to a Facebook employee involved in planning the meeting, part of the goal was to bring in a group of conservatives who were certain to fight with one another. They made sure to have libertarians who wouldn’t want to regulate the platform and partisans who would. Another goal, according to the employee, was to make sure the attendees were “bored to death” by a technical presentation after Zuckerberg and Sandberg had addressed the group.

The power went out, and the room got uncomfortably hot. But otherwise the meeting went according to plan. The guests did indeed fight, and they failed to unify in a way that was either threatening or coherent. Some wanted the company to set hiring quotas for conservative employees; others thought that idea was nuts. As often happens when outsiders meet with Facebook, people used the time to try to figure out how they could get more followers for their own pages.

Afterward, Glenn Beck, one of the invitees, wrote an essay about the meeting, praising Zuckerberg. “I asked him if Facebook, now or in the future, would be an open platform for the sharing of all ideas or a curator of content,” Beck wrote. “Without hesitation, with clarity and boldness, Mark said there is only one Facebook and one path forward: ‘We are an open platform.’”

Inside Facebook itself, the backlash around Trending Topics did inspire some genuine soul-searching. But none of it got very far. A quiet internal project, codenamed Hudson, cropped up around this time to determine, according to someone who worked on it, whether News Feed should be modified to better deal with some of the most complex issues facing the product. Does it favor posts that make people angry? Does it favor simple or even false ideas over complex and true ones? Those are hard questions, and the company didn’t have answers to them yet. Ultimately, in late June, Facebook announced a modest change: The algorithm would be revised to favor posts from friends and family. At the same time, Adam Mosseri, Facebook’s News Feed boss, posted a manifesto titled “Building a Better News Feed for You.” People inside Facebook spoke of it as a document roughly resembling the Magna Carta; the company had never spoken before about how News Feed really worked. To outsiders, though, the document came across as boilerplate. It said roughly what you’d expect: that the company was opposed to clickbait but that it wasn’t in the business of favoring certain kinds of viewpoints.

The most important consequence of the Trending Topics controversy, according to nearly a dozen former and current employees, was that Facebook became wary of doing anything that might look like stifling conservative news. It had burned its fingers once and didn’t want to do it again. And so a summer of deeply partisan rancor and calumny began with Facebook eager to stay out of the fray.

IV

Facebook thought that News Corp was threatening to push for a government antitrust investigation or maybe an inquiry into whether the company deserved its protection from liability as a neutral platform. Inside Facebook, executives believed Murdoch might use his papers and TV stations to amplify critiques of the company. News Corp says that was not at all the case; the company threatened to deploy executives, but not its journalists.

Zuckerberg had reason to take the meeting especially seriously, according to a former Facebook executive, because he had firsthand knowledge of Murdoch’s skill in the dark arts. Back in 2007, Facebook had come under criticism from 49 state attorneys general for failing to protect young Facebook users from sexual predators and inappropriate content. Concerned parents had written to Connecticut attorney general Richard Blumenthal, who opened an investigation, and to The New York Times, which published a story. But according to a former Facebook executive in a position to know, the company believed that many of the Facebook accounts and the predatory behavior the letters referenced were fakes, traceable to News Corp lawyers or others working for Murdoch, who owned Facebook’s biggest competitor, MySpace. “We traced the creation of the Facebook accounts to IP addresses at the Apple store a block away from the MySpace offices in Santa Monica,” the executive says. “Facebook then traced interactions with those accounts to News Corp lawyers. When it comes to Facebook, Murdoch has been playing every angle he can for a long time.” (Both News Corp and its spinoff 21st Century Fox declined to comment.)

When Zuckerberg returned from Sun Valley, he told his employees that things had to change. They still weren’t in the news business, but they had to make sure there would be a news business. And they had to communicate better. One of those who got a new to-do list was Andrew Anker, a product manager who’d arrived at Facebook in 2015 after a career in journalism (including a long stint at WIRED in the ’90s). One of his jobs was to help the company think through how publishers could make money on the platform. Shortly after Sun Valley, Anker met with Zuckerberg and asked to hire 60 new people to work on partnerships with the news industry. Before the meeting ended, the request was approved.

But having more people out talking to publishers just drove home how hard it would be to resolve the financial problems Murdoch wanted fixed. News outfits were spending millions to produce stories that Facebook was benefiting from, and Facebook, they felt, was giving too little back in return. Instant Articles, in particular, struck them as a Trojan horse. Publishers complained that they could make more money from stories that loaded on their own mobile web pages than on Facebook Instant. (They often did so, it turned out, in ways that short-changed advertisers, by sneaking in ads that readers were unlikely to see. Facebook didn’t let them get away with that.) Another seemingly irreconcilable difference: Outlets like Murdoch’s Wall Street Journal depended on paywalls to make money, but Instant Articles banned paywalls; Zuckerberg disapproved of them. After all, he would often ask, how exactly do walls and toll booths make the world more open and connected?

The conversations often ended at an impasse, but Facebook was at least becoming more attentive. This newfound appreciation for the concerns of journalists did not, however, extend to the journalists on Facebook’s own Trending Topics team. In late August, everyone on the team was told that their jobs were being eliminated. Simultaneously, authority over the algorithm shifted to a team of engineers based in Seattle. Very quickly the module started to surface lies and fiction. A headline days later read, “Fox News Exposes Traitor Megyn Kelly, Kicks Her Out For Backing Hillary.”

V

In the summer of 2016, at the top of the general election campaign, Trump’s digital operation might have seemed to be at a major disadvantage. After all, Hillary Clinton’s team was flush with elite talent and got advice from Eric Schmidt, known for running ­Google. Trump’s was run by Brad Parscale, known for setting up the Eric Trump Foundation’s web page. Trump’s social media director was his former caddie. But in 2016, it turned out you didn’t need digital experience running a presidential campaign, you just needed a knack for Facebook.

Over the course of the summer, Trump’s team turned the platform into one of its primary vehicles for fund-­raising. The campaign uploaded its voter files—the names, addresses, voting history, and any other information it had on potential voters—to Facebook. Then, using a tool called Look­alike Audiences, Facebook identified the broad characteristics of, say, people who had signed up for Trump newsletters or bought Trump hats. That allowed the campaign to send ads to people with similar traits. Trump would post simple messages like “This election is being rigged by the media pushing false and unsubstantiated charges, and outright lies, in order to elect Crooked Hillary!” that got hundreds of thousands of likes, comments, and shares. The money rolled in. Clinton’s wonkier messages, meanwhile, resonated less on the platform. Inside Facebook, almost everyone on the executive team wanted Clinton to win; but they knew that Trump was using the platform better. If he was the candidate for Facebook, she was the candidate for LinkedIn.

Trump’s candidacy also proved to be a wonderful tool for a new class of scammers pumping out massively viral and entirely fake stories. Through trial and error, they learned that memes praising the former host of The Apprentice got many more readers than ones praising the former secretary of state. A website called Ending the Fed proclaimed that the Pope had endorsed Trump and got almost a million comments, shares, and reactions on Facebook, according to an analysis by BuzzFeed. Other stories asserted that the former first lady had quietly been selling weapons to ISIS, and that an FBI agent suspected of leaking Clinton’s emails was found dead. Some of the posts came from hyperpartisan Americans. Some came from overseas content mills that were in it purely for the ad dollars. By the end of the campaign, the top fake stories on the platform were generating more engagement than the top real ones.

Even current Facebookers acknowledge now that they missed what should have been obvious signs of people misusing the platform. And looking back, it’s easy to put together a long list of possible explanations for the myopia in Menlo Park about fake news. Management was gun-shy because of the Trending Topics fiasco; taking action against partisan disinformation—or even identifying it as such—might have been seen as another act of political favoritism. Facebook also sold ads against the stories, and sensational garbage was good at pulling people into the platform. Employees’ bonuses can be based largely on whether Facebook hits certain growth and revenue targets, which gives people an extra incentive not to worry too much about things that are otherwise good for engagement. And then there was the ever-present issue of Section 230 of the 1996 Communications Decency Act. If the company started taking responsibility for fake news, it might have to take responsibility for a lot more. Facebook had plenty of reasons to keep its head in the sand.

Roger McNamee, however, watched carefully as the nonsense spread. First there were the fake stories pushing Bernie Sanders, then he saw ones supporting Brexit, and then helping Trump. By the end of the summer, he had resolved to write an op-ed about the problems on the platform. But he never ran it. “The idea was, look, these are my friends. I really want to help them.” And so on a Sunday evening, nine days before the 2016 election, McNamee emailed a 1,000-word letter to Sandberg and Zuckerberg. “I am really sad about Facebook,” it began. “I got involved with the company more than a decade ago and have taken great pride and joy in the company’s success … until the past few months. Now I am disappointed. I am embarrassed. I am ashamed.”

VI

Zuckerberg declined to be interviewed for this article, but people who know him well say he likes to form his opinions from data. And in this case he wasn’t without it. Before the interview, his staff had worked up a back-of-the-­envelope calculation showing that fake news was a tiny percentage of the total amount of election-­related content on the platform. But the analysis was just an aggregate look at the percentage of clearly fake stories that appeared across all of Facebook. It didn’t measure their influence or the way fake news affected specific groups. It was a number, but not a particularly meaningful one.

Zuckerberg’s comments did not go over well, even inside Facebook. They seemed clueless and self-absorbed. “What he said was incredibly damaging,” a former executive told WIRED. “We had to really flip him on that. We realized that if we didn’t, the company was going to start heading down this pariah path that Uber was on.”

A week after his “pretty crazy” comment, Zuckerberg flew to Peru to give a talk to world leaders about the ways that connecting more people to the internet, and to Facebook, could reduce global poverty. Right after he landed in Lima, he posted something of a mea culpa. He explained that Facebook did take misinformation seriously, and he presented a vague seven-point plan to tackle it. When a professor at the New School named David Carroll saw Zuckerberg’s post, he took a screenshot. Alongside it on Carroll’s feed ran a headline from a fake CNN with an image of a distressed Donald Trump and the text “DISQUALIFIED; He’s GONE!”

At the conference in Peru, Zuckerberg met with a man who knows a few things about politics: Barack Obama. Media reports portrayed the encounter as one in which the lame-duck president pulled Zuckerberg aside and gave him a “wake-up call” about fake news. But according to someone who was with them in Lima, it was Zuckerberg who called the meeting, and his agenda was merely to convince Obama that, yes, Facebook was serious about dealing with the problem. He truly wanted to thwart misinformation, he said, but it wasn’t an easy issue to solve.

Meanwhile, at Facebook, the gears churned. For the first time, insiders really began to question whether they had too much power. One employee told WIRED that, watching Zuckerberg, he was reminded of Lennie in Of Mice and Men, the farm-worker with no understanding of his own strength.

Very soon after the election, a team of employees started working on something called the News Feed Integrity Task Force, inspired by a sense, one of them told WIRED, that hyperpartisan misinformation was “a disease that’s creeping into the entire platform.” The group, which included Mosseri and Anker, began to meet every day, using whiteboards to outline different ways they could respond to the fake-news crisis. Within a few weeks the company announced it would cut off advertising revenue for ad farms and make it easier for users to flag stories they thought false.

In December the company announced that, for the first time, it would introduce fact-checking onto the platform. Facebook didn’t want to check facts itself; instead it would outsource the problem to professionals. If Facebook received enough signals that a story was false, it would automatically be sent to partners, like Snopes, for review. Then, in early January, Facebook announced that it had hired Campbell Brown, a former anchor at CNN. She immediately became the most prominent journalist hired by the company.

Soon Brown was put in charge of something called the Facebook Journalism Project. “We spun it up over the holidays, essentially,” says one person involved in discussions about the project. The aim was to demonstrate that Facebook was thinking hard about its role in the future of journalism—essentially, it was a more public and organized version of the efforts the company had begun after Murdoch’s tongue-lashing. But sheer anxiety was also part of the motivation. “After the election, because Trump won, the media put a ton of attention on fake news and just started hammering us. People started panicking and getting afraid that regulation was coming. So the team looked at what Google had been doing for years with News Lab”—a group inside Alphabet that builds tools for journalists—“and we decided to figure out how we could put together our own packaged program that shows how seriously we take the future of news.”

Facebook was reluctant, however, to issue any mea culpas or action plans with regard to the problem of filter bubbles or Facebook’s noted propensity to serve as a tool for amplifying outrage. Members of the leadership team regarded these as issues that couldn’t be solved, and maybe even shouldn’t be solved. Was Facebook really more at fault for amplifying outrage during the election than, say, Fox News or MSNBC? Sure, you could put stories into people’s feeds that contradicted their political viewpoints, but people would turn away from them, just as surely as they’d flip the dial back if their TV quietly switched them from Sean Hannity to Joy Reid. The problem, as Anker puts it, “is not Facebook. It’s humans.”

VII

Roger McNamee, meanwhile, was getting steamed at Facebook’s response to his letter. Zuckerberg and Sandberg had written him back promptly, but they hadn’t said anything substantial. Instead he ended up having a months-long, ultimately futile set of email exchanges with Dan Rose, Facebook’s VP for partnerships. McNamee says Rose’s message was polite but also very firm: The company was doing a lot of good work that McNamee couldn’t see, and in any event Facebook was a platform, not a media company.

“And I’m sitting there going, ‘Guys, seriously, I don’t think that’s how it works,’” McNamee says. “You can assert till you’re blue in the face that you’re a platform, but if your users take a different point of view, it doesn’t matter what you assert.”

As the saying goes, heaven has no rage like love to hatred turned, and McNamee’s concern soon became a cause—and the beginning of an alliance. In April 2017 he connected with a former Google design ethicist named Tristan Harris when they appeared together on Bloomberg TV. Harris had by then gained a national reputation as the conscience of Silicon Valley. He had been profiled on 60 Minutes and in The Atlantic, and he spoke eloquently about the subtle tricks that social media companies use to foster an addiction to their services. “They can amplify the worst aspects of human nature,” Harris told WIRED this past December. After the TV appearance, McNamee says he called Harris up and asked, “Dude, do you need a wingman?”

The next month, DiResta published an ­article comparing purveyors of disinformation on social media to manipulative high-frequency traders in financial markets. “Social networks enable malicious actors to operate at platform scale, because they were designed for fast information flows and virality,” she wrote. Bots and sock puppets could cheaply “create the illusion of a mass groundswell of grassroots activity,” in much the same way that early, now-illegal trading algorithms could spoof demand for a stock. Harris read the article, was impressed, and emailed her.

The three were soon out talking to anyone who would listen about Facebook’s poisonous effects on American democracy. And before long they found receptive audiences in the media and Congress—groups with their own mounting grievances against the social media giant.

VIII

And then there’s the simple, deep fear and mistrust that Facebook inspires. Every publisher knows that, at best, they are sharecroppers on Facebook’s massive industrial farm. The social network is roughly 200 times more valuable than the Times. And journalists know that the man who owns the farm has the leverage. If Facebook wanted to, it could quietly turn any number of dials that would harm a publisher—by manipulating its traffic, its ad network, or its readers.

Emissaries from Facebook, for their part, find it tiresome to be lectured by people who can’t tell an algorithm from an API. They also know that Facebook didn’t win the digital ad market through luck: It built a better ad product. And in their darkest moments, they wonder: What’s the point? News makes up only about 5 percent of the total content that people see on Facebook globally. The company could let it all go and its shareholders would scarcely notice. And there’s another, deeper problem: Mark Zuckerberg, according to people who know him, prefers to think about the future. He’s less interested in the news industry’s problems right now; he’s interested in the problems five or 20 years from now. The editors of major media companies, on the other hand, are worried about their next quarter—maybe even their next phone call. When they bring lunch back to their desks, they know not to buy green bananas.

This mutual wariness—sharpened almost to enmity in the wake of the election—did not make life easy for Campbell Brown when she started her new job running the nascent Facebook Journalism Project. The first item on her to-do list was to head out on yet another Facebook listening tour with editors and publishers. One editor describes a fairly typical meeting: Brown and Chris Cox, Facebook’s chief product officer, invited a group of media leaders to gather in late January 2017 at Brown’s apartment in Manhattan. Cox, a quiet, suave man, sometimes referred to as “the Ryan Gosling of Facebook Product,” took the brunt of the ensuing abuse. “Basically, a bunch of us just laid into him about how Facebook was destroying journalism, and he graciously absorbed it,” the editor says. “He didn’t much try to defend them. I think the point was really to show up and seem to be listening.” Other meetings were even more tense, with the occasional comment from journalists noting their interest in digital antitrust issues.

As bruising as all this was, Brown’s team became more confident that their efforts were valued within the company when Zuckerberg published a 5,700-word corporate manifesto in February. He had spent the previous three months, according to people who know him, contemplating whether he had created something that did more harm than good. “Are we building the world we all want?” he asked at the beginning of his post, implying that the answer was an obvious no. Amid sweeping remarks about “building a global community,” he emphasized the need to keep people informed and to knock out false news and clickbait. Brown and others at Facebook saw the manifesto as a sign that Zuckerberg understood the company’s profound civic responsibilities. Others saw the document as blandly grandiose, showcasing Zuckerberg’s tendency to suggest that the answer to nearly any problem is for people to use Facebook more.

Shortly after issuing the manifesto, Zuckerberg set off on a carefully scripted listening tour of the country. He began popping into candy shops and dining rooms in red states, camera crew and personal social media team in tow. He wrote an earnest post about what he was learning, and he deflected questions about whether his real goal was to become president. It seemed like a well-­meaning effort to win friends for Facebook. But it soon became clear that Facebook’s biggest problems emanated from places farther away than Ohio.

IX

That holy shit moment, though, didn’t come until more than six months after the election. Early in the campaign season, Facebook was aware of familiar attacks emanating from known Russian hackers, such as the group APT28, which is believed to be affiliated with Moscow. They were hacking into accounts outside of Facebook, stealing documents, then creating fake Facebook accounts under the banner of DCLeaks, to get people to discuss what they’d stolen. The company saw no signs of a serious, concerted foreign propaganda campaign, but it also didn’t think to look for one.

During the spring of 2017, the company’s security team began preparing a report about how Russian and other foreign intelligence operations had used the platform. One of its authors was Alex Stamos, head of Facebook’s security team. Stamos was something of an icon in the tech world for having reportedly resigned from his previous job at Yahoo after a conflict over whether to grant a US intelligence agency access to Yahoo servers. According to two people with direct knowledge of the document, he was eager to publish a detailed, specific analysis of what the company had found. But members of the policy and communications team pushed back and cut his report way down. Sources close to the security team suggest the company didn’t want to get caught up in the political whirlwind of the moment. (Sources on the politics and communications teams insist they edited the report down, just because the darn thing was hard to read.)

On April 27, 2017, the day after the Senate announced it was calling then FBI director James Comey to testify about the Russia investigation, Stamos’ report came out. It was titled “Information Operations and Facebook,” and it gave a careful step-by-step explanation of how a foreign adversary could use Facebook to manipulate people. But there were few specific examples or details, and there was no direct mention of Russia. It felt bland and cautious. As Renée DiResta says, “I remember seeing the report come out and thinking, ‘Oh, goodness, is this the best they could do in six months?’”

One month later, a story in Time suggested to Stamos’ team that they might have missed something in their analysis. The article quoted an unnamed senior intelligence official saying that Russian operatives had bought ads on Facebook to target Americans with propaganda. Around the same time, the security team also picked up hints from congressional investigators that made them think an intelligence agency was indeed looking into Russian Facebook ads. Caught off guard, the team members started to dig into the company’s archival ads data themselves.

Eventually, by sorting transactions according to a series of data points—Were ads purchased in rubles? Were they purchased within browsers whose language was set to Russian?—they were able to find a cluster of accounts, funded by a shadowy Russian group called the Internet Research Agency, that had been designed to manipulate political opinion in America. There was, for example, a page called Heart of Texas, which pushed for the secession of the Lone Star State. And there was Blacktivist, which pushed stories about police brutality against black men and women and had more followers than the verified Black Lives Matter page.

Numerous security researchers express consternation that it took Facebook so long to realize how the Russian troll farm was exploiting the platform. After all, the group was well known to Facebook. Executives at the company say they’re embarrassed by how long it took them to find the fake accounts, but they point out that they were never given help by US intelligence agencies. A staffer on the Senate Intelligence Committee likewise voiced exasperation with the company. “It seemed obvious that it was a tactic the Russians would exploit,” the staffer says.

When Facebook finally did find the Russian propaganda on its platform, the discovery set off a crisis, a scramble, and a great deal of confusion. First, due to a miscalculation, word initially spread through the company that the Russian group had spent millions of dollars on ads, when the actual total was in the low six figures. Once that error was resolved, a disagreement broke out over how much to reveal, and to whom. The company could release the data about the ads to the public, release everything to Congress, or release nothing. Much of the argument hinged on questions of user privacy. Members of the security team worried that the legal process involved in handing over private user data, even if it belonged to a Russian troll farm, would open the door for governments to seize data from other Facebook users later on. “There was a real debate internally,” says one executive. “Should we just say ‘Fuck it’ and not worry?” But eventually the company decided it would be crazy to throw legal caution to the wind “just because Rachel Maddow wanted us to.”

Ultimately, a blog post appeared under Stamos’ name in early September announcing that, as far as the company could tell, the Russians had paid Facebook $100,000 for roughly 3,000 ads aimed at influencing American politics around the time of the 2016 election. Every sentence in the post seemed to downplay the substance of these new revelations: The number of ads was small, the expense was small. And Facebook wasn’t going to release them. The public wouldn’t know what they looked like or what they were really aimed at doing.

This didn’t sit at all well with DiResta. She had long felt that Facebook was insufficiently forthcoming, and now it seemed to be flat-out stonewalling. “That was when it went from incompetence to malice,” she says. A couple of weeks later, while waiting at a Walgreens to pick up a prescription for one of her kids, she got a call from a researcher at the Tow Center for Digital Journalism named Jonathan Albright. He had been mapping ecosystems of misinformation since the election, and he had some excellent news. “I found this thing,” he said. Albright had started digging into CrowdTangle, one of the analytics platforms that Facebook uses. And he had discovered that the data from six of the accounts Facebook had shut down were still there, frozen in a state of suspended animation. There were the posts pushing for Texas secession and playing on racial antipathy. And then there were political posts, like one that referred to Clinton as “that murderous anti-American traitor Killary.” Right before the election, the Blacktivist account urged its supporters to stay away from Clinton and instead vote for Jill Stein. Albright downloaded the most recent 500 posts from each of the six groups. He reported that, in total, their posts had been shared more than 340 million times.

X

McNamee and Harris had first traveled to DC for a day in July to meet with members of Congress. Then, in September, they were joined by DiResta and began spending all their free time counseling senators, representatives, and members of their staffs. The House and Senate Intelligence Committees were about to hold hearings on Russia’s use of social media to interfere in the US election, and McNamee, Harris, and ­DiResta were helping them prepare. One of the early questions they weighed in on was the matter of who should be summoned to testify. Harris recommended that the CEOs of the big tech companies be called in, to create a dramatic scene in which they all stood in a neat row swearing an oath with their right hands in the air, roughly the way tobacco executives had been forced to do a generation earlier. Ultimately, though, it was determined that the general counsels of the three companies—Facebook, Twitter, and Google—should head into the lion’s den.

And so on November 1, Colin Stretch arrived from Facebook to be pummeled. During the hearings themselves, DiResta was sitting on her bed in San Francisco, watching them with her headphones on, trying not to wake up her small children. She listened to the back-and-forth in Washington while chatting on Slack with other security researchers. She watched as Marco Rubio smartly asked whether Facebook even had a policy forbidding foreign governments from running an influence campaign through the platform. The answer was no. Rhode Island senator Jack Reed then asked whether Facebook felt an obligation to individually notify all the users who had seen Russian ads that they had been deceived. The answer again was no. But maybe the most threatening comment came from Dianne Feinstein, the senior senator from Facebook’s home state. “You’ve created these platforms, and now they’re being misused, and you have to be the ones to do something about it,” she declared. “Or we will.”

After the hearings, yet another dam seemed to break, and former Facebook executives started to go public with their criticisms of the company too. On November 8, billionaire entrepreneur Sean Parker, Facebook’s first president, said he now regretted pushing Facebook so hard on the world. “I don’t know if I really understood the consequences of what I was saying,” he said. “God only knows what it’s doing to our children’s brains.” Eleven days later, Facebook’s former privacy manager, Sandy Parakilas, published a New York Times op-ed calling for the government to regulate Facebook: “The company won’t protect us by itself, and nothing less than our democracy is at stake.”

XI

Other signs emerged, too, that Zuckerberg was beginning to absorb the criticisms of his company. The Facebook Journalism Project, for instance, seemed to be making the company take its obligations as a publisher, and not just a platform, more seriously. In the fall, the company announced that Zuckerberg had decided—after years of resisting the idea—that publishers using Facebook Instant Articles could require readers to subscribe. Paying for serious publications, in the months since the election, had come to seem like both the path forward for journalism and a way of resisting the post-truth political landscape. (WIRED recently instituted its own paywall.) Plus, offering subscriptions arguably helped put in place the kinds of incentives that Zuckerberg professed to want driving the platform. People like Alex Hardiman, the head of Facebook news products and an alum of The New York Times, started to recognize that Facebook had long helped to create an economic system that rewarded publishers for sensationalism, not accuracy or depth. “If we just reward content based on raw clicks and engagement, we might actually see content that is increasingly sensationalist, clickbaity, polarizing, and divisive,” she says. A social network that rewards only clicks, not subscriptions, is like a dating service that encourages one-night stands but not marriages.

XII

During the late fall, criticism continued to mount: Facebook was accused of becoming a central vector for spreading deadly propaganda against the Rohingya in Myanmar and for propping up the brutal leadership of Rodrigo Duterte in the Philippines. And December brought another haymaker from someone closer by. Early that month, it emerged that Chamath Palihapitiya, who had been Facebook’s vice president for user growth before leaving in 2011, had told an audience at Stanford that he thought social media platforms like Facebook had “created tools that are ripping apart the social fabric” and that he feels “tremendous guilt” for being part of that. He said he tries to use Facebook as little as possible and doesn’t permit his children to use such platforms at all.

The criticism stung in a way that others hadn’t. Palihapitiya is close to many of the top executives at Facebook, and he has deep cachet in Silicon Valley and among Facebook engineers as a part-owner of the Golden State Warriors. Sheryl Sandberg sometimes wears a chain around her neck that’s welded together from one given to her by Zuckerberg and one given to her by Palihapitiya after her husband’s death. The company issued a statement saying it had been a long time since Palihapitiya had worked there. “Facebook was a very different company back then and as we have grown we have realized how our responsibilities have grown too.” Asked why the company had responded to Palihapitiya, and not to others, a senior Facebook executive said, “Chamath is—was—a friend to a lot of people here.”

Roger McNamee, meanwhile, went on a media tour lambasting the company. He published an essay in Washington Monthly and then followed up in The Washington Post and The Guardian. Facebook was less impressed with him. Executives considered him to be overstating his connection to the company and dining out on his criticism. Andrew Bos­worth, a VP and member of the management team, tweeted, “I’ve worked at Facebook for 12 years and I have to ask: Who the fuck is Roger McNamee?”

Zuckerberg did seem to be eager to mend one fence, though. Around this time, a team of Facebook executives gathered for dinner with executives from News Corp at the Grill, an upscale restaurant in Manhattan. Right at the start, Zuckerberg raised a toast to Murdoch. He spoke charmingly about reading a biography of the older man and of admiring his accomplishments. Then he described a game of tennis he’d once played against Murdoch. At first he had thought it would be easy to hit the ball with a man more than 50 years his senior. But he quickly realized, he said, that Murdoch was there to compete.

XIII

That New Year’s challenge, it turned out, was a bit of carefully considered choreography setting up a series of announcements, starting with a declaration the following week that the News Feed algorithm would be rejiggered to favor “meaningful interactions.” Posts and videos of the sort that make us look or like—but not comment or care—would be deprioritized. The idea, explained Adam Mosseri, is that, online, “interacting with people is positively correlated with a lot of measures of well-being, whereas passively consuming content online is less so.”

To numerous people at the company, the announcement marked a huge departure. Facebook was putting a car in reverse that had been driving at full speed in one direction for 14 years. Since the beginning, Zuckerberg’s ambition had been to create another internet, or perhaps another world, inside of Facebook, and to get people to use it as much as possible. The business model was based on advertising, and advertising was insatiably hungry for people’s time. But now Zuckerberg said he expected these new changes to News Feed would make people use Facebook less.

The announcement was hammered by many in the press. During the rollout, Mosseri explained that Facebook would downgrade stories shared by businesses, celebrities, and publishers, and prioritize stories shared by friends and family. Critics surmised that these changes were just a way of finally giving the publishing industry a middle finger. “Facebook has essentially told media to kiss off,” Franklin Foer wrote in The Atlantic. “Facebook will be back primarily in the business of making us feel terrible about the inferiority of our vacations, the relative mediocrity of our children, teasing us into sharing more of our private selves.”

But inside Facebook, executives insist this isn’t remotely the case. According to Anker, who retired from the company in December but worked on these changes, and who has great affection for the management team, “It would be a mistake to see this as a retreat from the news industry. This is a retreat from ‘Anything goes if it works with our algorithm to drive up engagement.’” According to others still at the company, Zuckerberg didn’t want to pull back from actual journalism. He just genuinely wanted there to be less crap on the platform: fewer stories with no substance; fewer videos you can watch without thinking.

And then, a week after telling the world about “meaningful interactions,” Zuckerberg announced another change that seemed to answer these concerns, after a fashion. For the first time in the company’s history, he said in a note posted to his personal page, Facebook will start to boost certain publishers—ones whose content is “trustworthy, informative, and local.” For the past year, Facebook has been developing algorithms to hammer publishers whose content is fake; now it’s trying to elevate what’s good. For starters, he explained, the company would use reader surveys to determine which sources are trustworthy. That system, critics were quick to point out, will surely be gamed, and many people will say they trust sources just because they recognize them. But this announcement, at least, went over a little better in boardrooms and newsrooms. Right after the post went up, the stock price of The New York Times shot up—as did that of News Corp.

Zuckerberg has hinted—and insiders have confirmed—that we should expect a year of more announcements like this. The company is experimenting with giving publishers more control over paywalls and allowing them to feature their logos more prominently to reestablish the brand identities that Facebook flattened years ago. One somewhat hostile outside suggestion has come from Facebook’s old antagonist Murdoch, who said in late January that if Facebook truly valued “trustworthy” publishers, it should pay them carriage fees.

The fate that Facebook really cares about, however, is its own. It was built on the power of network effects: You joined because everyone else was joining. But network effects can be just as powerful in driving people off a platform. Zuckerberg understands this viscerally. After all, he helped create those problems for MySpace a decade ago and is arguably doing the same to Snap today. Zuckerberg has avoided that fate, in part, because he has proven brilliant at co-opting his biggest threats. When social media started becoming driven by images, he bought Instagram. When messaging took off, he bought WhatsApp. When Snapchat became a threat, he copied it. Now, with all his talk of “time well spent,” it seems as if he’s trying to co-opt Tristan Harris too.

But people who know him say that Zuckerberg has truly been altered in the crucible of the past several months. He has thought deeply; he has reckoned with what happened; and he truly cares that his company fix the problems swirling around it. And he’s also worried. “This whole year has massively changed his personal techno-­optimism,” says an executive at the company. “It has made him much more paranoid about the ways that people could abuse the thing that he built.”

The past year has also altered Facebook’s fundamental understanding about whether it’s a publisher or a platform. The company has always answered that question defiantly—platform, platform, platform—for regulatory, financial, and maybe even emotional reasons. But now, gradually, Facebook has evolved. Of course it’s a platform, and always will be. But the company also realizes now that it bears some of the responsibilities that a publisher does: for the care of its readers, and for the care of the truth. You can’t make the world more open and connected if you’re breaking it apart. So what is it: publisher or platform? Facebook seems to have finally recognized that it is quite clearly both.