PDF Summary:Careless People, by Sarah Wynn-Williams

Can tech leaders be trusted to shape our future? What happens when profit and power trump ethics and accountability? In Careless People, Sarah Wynn-Williams—Facebook’s Director of Global Policy from 2011-2017—offers an insider account of her time working alongside Mark Zuckerberg and other Facebook leaders. She was an early believer in the company’s potential to connect the world, but she became disillusioned after its leadership evolved to prioritize profit over responsibility.

In this memoir and tell-all, Wynn-Williams alleges that Facebook’s top leaders were dangerously reckless people who caused great harm in the world. You’ll learn in this guide how Facebook’s advertising tools helped shape the 2016 US presidential election, how the company’s negligence led to violence in Myanmar, and what their collaboration with China’s Communist Party means for the future of AI. Whether you’re concerned about social media’s influence on politics, interested in ethical leadership, or curious how to prevent tech companies from causing more harm, our guide provides insights into one of the most powerful companies in history.

(continued)...

This may explain why Sandberg didn’t take action when the Feminist Fight Club brought up the issue of systemic harassment at Facebook: Perhaps she didn’t believe in (or see any benefit in) systemic change. Feminist critic bell hooks points this out, arguing that Sandberg takes a simplistic and insufficient approach to women’s empowerment. hooks adds that this approach fails to address broader issues of race, class, and privilege. She says that Sandberg herself benefits from privilege as a wealthy, educated white woman, so she’s out of touch with the issues that women of color face—like systemic barriers that individual effort can’t overcome.

In a 2017 Facebook post, Sandberg acknowledged the need for “systemic, lasting changes that deter bad behavior and protect everyone,” saying that she’d experienced some amount of harassment herself. But per Wynn-Williams, Sandberg didn’t challenge such structures at Facebook—which underscores hooks’ points—and in rising up the ladder, Sandberg may have become like the men already at the top.

Sandberg’s Politicking

Wynn-Williams also writes that Sandberg was a ruthless political actor who played a large role in deflecting criticism from Facebook and managing its reputation. For instance, when Facebook was criticized for doing too little to combat misinformation and its political consequences in Myanmar (an issue we’ll discuss later), Sandberg downplayed Facebook’s responsibility for the problem and exaggerated how much they were doing to address it.

(Shortform note: Though Wynn-Williams criticizes Sandberg’s focus on image management, Sandberg seems to have taken responsibility for at least one key issue. During 2018 Senate Intelligence Committee hearings, Sandberg testified alongside Twitter CEO Jack Dorsey about Russian interference in the 2016 US presidential election. She acknowledged, as the Committee found, that actors connected to Russia used Facebook to sow social discord by spreading misinformation about politically sensitive topics, like race and immigration, to help Trump win. She also said Facebook had been “too slow” to respond to the issue, and that they were investing in security—like more staff and AI solutions—to fight future interference.)

Kaplan: Biased and Badly Behaved

A few years into Wynn-Williams’ work at Facebook, Joel Kaplan became her manager. Wynn-Williams says Kaplan was a savvy and well-connected political operative who’d been an adviser to US president George Bush and maintained ties to the Republican Party. He wasn’t as publicly visible as Zuckerberg or Sandberg, but he had major influence over Facebook’s approach to politics, policy, and content moderation.

According to Wynn-Williams, Kaplan ensured that Facebook didn’t moderate conservative content, however extreme or hateful it got. She says this enabled the spread of misinformation and disinformation on the platform. She also alleges that Kaplan behaved inappropriately toward her on multiple occasions.

(Shortform note: Kaplan’s move from the Bush White House to Facebook exemplifies the “revolving door” phenomenon—when tech lobbyists become political staffers (and vice versa). Political figures such as US Senator Elizabeth Warren have criticized the practice, suggesting the hires are motivated by corporate political agendas. In 2019, Warren pointed out that after hiring Kaplan, Facebook’s lobbying spend increased 100-fold to over $71 million. What did that money buy them? If Warren and Wynn-Williams are to be believed, that money got Facebook closed-door meetings to influence US politicians. This may explain why Facebook condoned certain hate speech—moderating it more might’ve lost them key relationships on Capitol Hill.)

Kaplan’s Conservative Bias

Wynn-Williams writes that Kaplan often pushed for internal decisions that would benefit right-wing content and personalities. For instance, when Facebook was facing scrutiny from conservative lawmakers because of its apparent liberal bias, Kaplan ensured that Facebook didn’t censor conservative pages that had begun spreading hate speech and disinformation. Wynn-Williams makes no judgment on his politics per se, but she does suggest that his influence—especially around the 2016 US election—was a big part of why Facebook did so little to combat political disinformation.

(Shortform note: When does free expression cross into harmful territory? While social media platforms aren't bound by the First Amendment, which prohibits government censorship of speech, they wield enormous influence over public conversation. So the question is: Where is the line between censorship and necessary moderation? One argument could be that Kaplan’s favoritism for conservative content wasn’t actually promoting free speech—it was selectively protecting certain voices in a way that created a toxic environment undermining the speech of others. Another argument could point out that hate speech is protected by the First Amendment. Media companies like Facebook sit in the difficult position of having to draw these lines, and they’ve drawn criticism for leaning either way.)

Kaplan’s Dismissiveness

According to Wynn-Williams, Kaplan also played a key role in Facebook’s morally fraught dealings with China and Myanmar (which we’ll cover in more detail later). In both cases, he ignored or dismissed red flags, such as violent rhetoric on Facebook in Myanmar, and Facebook’s willingness to build surveillance tools for China. Wynn-Williams says that his actions revealed his beliefs: Business growth and profit mattered more to him than ethics or human rights.

(Shortform note: Some might argue that Facebook’s leadership felt they were bound to maximize stakeholder value, a core responsibility for any company. Former Facebook investor Roger McNamee calls this a flimsy excuse, though, saying that ethical duty should come before profit. As it stands, business-as-usual capitalism tends to incentivize profit-seeking and doesn’t always account for externalities (the unintended consequences of business activities, like environmental damage or human rights violations). However, some argue businesses can become more ethical by expanding the definition of “stakeholder” to include the communities they operate in—and aiming to serve the people who live there in addition to just stockholders.)

Kaplan’s Harassment of Wynn-Williams

Beyond his power as a key decision-maker for Facebook’s political strategy, Kaplan allegedly used his position to sexually harass Wynn-Williams. She describes multiple occasions in which he treated her inappropriately, such as making suggestive comments about her body after childbirth and drunkenly grinding on her at a Facebook offsite event. These weren’t isolated incidents, she argues, but a consistent part of his character.

(Shortform note: According to federal US law, workplace sexual harassment is illegal under Title VII of the Civil Rights Act of 1964. It includes harassment like that Wynn-Williams describes—inappropriate touching and comments about someone’s appearance—as well as other types of harassment, such as making vulgar jokes and sending messages of a sexual nature. Title VII also describes the “remedies” available for victims of harassment who take legal action. These include compensation for emotional and/or physical distress, reinstatement if you lost your job, or a court order that your company change its policies to prevent future harassment.)

When Wynn-Williams eventually filed a formal complaint against him, the company launched an investigation. However, instead of holding Kaplan accountable, Wynn-Williams says Facebook’s investigation team cleared him before she had a chance to provide all her evidence (such as inappropriate emails). Kaplan had powerful allies, including Elliot Schrage (Facebook’s VP of public policy), who allegedly used his influence to ensure Kaplan wasn’t held accountable. Instead, Schrage fired Wynn-Williams on what she says were trumped-up charges—she was told she’d been underperforming—in the wake of her investigation filing.

(Shortform note: Wynn-Williams’ experience with Kaplan, as she describes it, follows common patterns displayed by many workplace harassment cases. According to data collected by the Equal Employment Opportunity Commission, over 75% of workplace harassment incidents involve a power dynamic where the harasser holds authority over the victim. This creates a situation where victims feel vulnerable and fear retaliation—retaliation that often happens, as in Wynn-Williams’ case.)

Facebook’s Harmful Impact

In the previous section, we explained how Wynn-Williams characterizes Facebook’s leaders during her time at the company. Next, we’ll explore the key events that she says happened as a result of their recklessness. These include Facebook’s role in genocide and political instability in Myanmar, its role in the 2016 US presidential election, and its collaboration with the Chinese Communist Party.

Facebook Enabled Violence in Myanmar

Wynn-Williams writes that in the mid-2010s, Facebook neglected to address hateful political rhetoric spreading on the platform in Myanmar. She says that this negligence led directly to real-world consequences: a violent campaign against the country’s Muslim Rohingya minority that the UN later called genocide. For years, Wynn-Williams and others repeatedly warned key decision-makers, but these leaders chose not to act—not because they didn’t know, but because they didn’t care.

(Shortform note: In December 2021, Rohingya refugees filed a $150 billion lawsuit against Meta in both US and UK courts, alleging the platform’s algorithms actively promoted hate speech and violence that contributed to genocide. The teams representing the Rohingya have charged that Facebook was negligent for allowing their engagement-focused algorithms to amplify inflammatory content on the platform. One attorney stated that “this genocide would not have happened without Facebook.” The US District Court of Northern California dismissed the case in 2023, saying the two-year statute of limitations had expired. Lawyers representing the Rohingya appealed that decision in late 2024, and the situation remains unresolved.)

Facebook’s unique role in Myanmar is central to understanding how this happened. Myanmar, formerly Burma, skipped over desktop computing and went straight to smartphones. Further, Facebook was so ubiquitous that most people thought it was the internet, or at least the main way to go online. This happened because Facebook had struck deals with telecoms to preload the app on mobile phones and not charge for data spent using it. All in all, the platform had unprecedented influence over how people in Myanmar connected, communicated, and shared information.

(Shortform note: Experts describe Myanmar’s jump straight to mobile computing as a case of “technological leapfrogging,” where a previously offline society skips desktop computing entirely. When Myanmar began opening up to the world in 2011, they rolled out mobile internet and rapidly brought the whole country online. Smartphone adoption rates soared. Facebook arrived in 2013 and launched its Internet.org initiative, providing free access to a limited number of websites, including Facebook itself. It quickly became the country’s most popular website. How might this sudden tech adoption have contributed to instability in the country? It’s hard to say for sure, but it’s possible that people weren’t prepared for how rapidly bad intentions could spread online.)

In 2015, Myanmar was preparing to hold its first democratic election in decades. But the military junta, reluctant to give up power, used Facebook to stir up chaos in the country and delay the process. Wynn-Williams’s team found that covert groups employed by the junta spread anti-Rohingya hate speech and nationalist propaganda on Facebook. Her team documented numerous inauthentic accounts being used to stoke division, impersonate influencers, and coordinate harassment. Meanwhile, peace activists and other good-faith actors were silenced or suppressed on the platform.

What was at first just online hate speech led to mobs rioting and burning mosques, shops, and the homes of Rohingya Muslims, according to Wynn-Williams. Despite this, Facebook’s leadership chose not to act. Further, they had only a single Burmese-speaking contractor on staff at the time—one person tasked with moderating Facebook in a country with tens of millions of users. Wynn-Williams managed to get a second Burmese speaker hired, but this was too little, too late.

In 2017, after Facebook’s leadership had ignored and dismissed the problems in Myanmar for years, the country’s junta launched a series of attacks against the Rohingya. Over 10,000 people were killed, thousands of women and girls were raped, villages were burned, and more than 700,000 Rohingya fled the country. The United Nations later concluded that Facebook had failed to moderate the hate speech and misinformation that led to this violence.

Social Media and Social Harm

Others agree that Facebook played a key role in the genesis of the Rohingya genocide. So does Facebook, which put out a 2018 press release stating they’d commissioned an independent review of the company’s conduct in Myanmar. The review found that Facebook hadn’t intervened enough, and the company acknowledged as much, saying it hadn’t done enough to prevent violence and would do more going forward.

However, a similar problem occurred again in Ethiopia from 2020-2022, when Facebook was again used to incite violence against ethnic minorities. Facebook had improved its conduct moderation systems by using AI flagging, but the platform still had far too few language experts on staff to moderate the platform. In 2021, they could monitor content in only four of the 85 languages spoken in Ethiopia. Internal memos record staffers trying to warn leadership of these problems as early as 2020, much as Wynn-Williams did for the Myanmar situation, but Facebook seemed to take action only in response to Frances Haugen’s leaking of the Facebook Papers in 2021.

The consequences in Ethiopia were devastating, if not on the same scale as those in Myanmar—for instance, some people belonging to a government-opposition political group were murdered by pro-government supporters.

Back in Myanmar, Facebook has made progress since 2017—expanding content moderation teams, developing better automated detection systems, and removing military-linked accounts. But hatred seems to be a problem that runs deeper than the platform that enables it to spread. Facebook has removed many pro-military propagandists, but they simply migrated to Telegram, which has far less moderation. These groups continue to incite violence against the people they hate. So while social media can amplify existing problems—like ethnic tensions and persecution—they may not be able to stamp them out altogether. Deeper solutions may require the coordinated efforts of peace groups, international governance, and tech companies.

Facebook’s Role in the 2016 US Presidential Election

Wynn-Williams also recounts how Facebook contributed to Donald Trump’s victory in the 2016 US presidential election. She writes that Trump’s team used Facebook’s powerful advertising tools to spread inflammatory and often counterfactual messaging on the platform—an approach which she views as unethical. Facebook leadership knew about this: They were selling campaign services (like help using their ad tools) to multiple candidates, including Trump and other Republicans. But they only realized they’d helped Trump win after the fact, according to Wynn-Williams, and by then it was too late.

(Shortform note: Some argue that during Trump’s second run for president, Facebook leaders explicitly aligned themselves with the Trump administration. They cite that in January 2025, Meta scrapped its fact-checking program, with Zuckerberg saying publicly that the platform’s previous efforts to moderate content were censorship. He’s announced that Facebook will follow the lead of X (formerly Twitter) in using crowd-sourced fact-checking, an approach championed by Elon Musk. This timing isn’t coincidental, some say—around the same time, Meta promoted Kaplan to lead policy, and Zuckerberg stated that the US was at a “cultural tipping point towards prioritizing speech.” Soon after, Trump praised the company, saying they’d “come a long way.”)

How did Facebook’s advertising tools contribute to Trump’s 2016 win, according to Wynn-Williams? Facebook ads allowed Trump’s campaign to target voters based on precise demographic data: age, gender, race, location (down to the ZIP code), political affiliation, income level, hobbies, and even emotional states. For example, Wynn-Williams describes a tactic in which Trump’s team targeted middle-aged Black men in cities like Philadelphia, who were likely to vote Democratic but had reservations about Hillary Clinton. The Trump team then sent these voters clips of her 1996 “superpredators” comment, widely seen as racist, to exploit those doubts and discourage voter turnout.

This worked so well because Facebook’s algorithm was optimized for engagement, and outrage strongly engages people. The more people Trump’s team could anger or rile up, the more people they could get to vote (or not to vote). Wynn-Williams says that Facebook leadership didn’t take this issue seriously because they saw a Trump victory as impossible. But in the meantime, he was a great customer for Facebook—the business profited massively from his campaign.

(Shortform note: According to one journalist, Facebook isn’t primarily a social media or advertising company—it's a surveillance giant. That is, Facebook is in the business of collecting as much data as they can about their users, and they’ve succeeded at building the most sophisticated surveillance systems in history. They track users’ every click, like, mouse movement, and keystroke are, and they sell access to that data in the form of the ad tools that the Trump campaign used. In this view, Facebook doesn’t combat misinformation because inflammatory content drives engagement (which has been found to be true for multiple platforms), generates more behavioral data for record and, in turn, raises their bottom line.)

Wynn-Williams contends that Facebook made no meaningful effort to curb the spread of Trump’s political disinformation. Zuckerberg and others downplayed the problem, despite internal warnings from employees and external pressure from researchers and journalists concerned about political manipulation. For his part, Kaplan made sure Facebook took no action that might be seen as anti-conservative, even when content crossed into hate speech or blatant falsehood.

After Trump’s victory, Facebook came under increasing scrutiny. But Zuckerberg needed to be convinced that Facebook had even played a role—Wynn-Williams says he was dismissive of that idea until Schrage (Facebook’s public policy VP and Wynn-Williams’ superior) explained to him in detail what had happened. Even then, Facebook’s leadership did little to address the platform’s friendliness to misleading or counterfactual content.

How Facebook’s AI Systems Struggle with Misinformation

Years on from the 2016 election, Facebook remains in a difficult spot when it comes to curbing the spread of misinformation and disinformation on the platform. For one, it’s very difficult (perhaps impossible) to fix the AI recommendation algorithms that fuel the problem. These algorithms determine which content people see on the platform. This is both because the technology is complex and because the way the algorithms work plays a key role in Facebook’s business success.

Facebook’s core recommendation engines work by maximizing user engagement. Internal charts show that the closer content gets to violating community standards, the more engagement it tends to drive. In other words, the same technology that grows Facebook’s profits also amplifies inflammatory content. This means that even when Facebook teams want to change the algorithm to reduce the spread of divisive content, they face pushback from others who point out that doing so would stall business growth.

Just as Zuckerberg needed convincing that Facebook had enabled Trump’s 2016 victory, Joaquin Quiñonero—creator of some of Facebook’s earliest recommendation algorithms—has expressed reluctance to acknowledge the effects of his work. When asked whether his systems have contributed to events like the 2021 US Capitol attack, Quiñonero said, "I don't know... That's my honest answer."

Facebook Collaborated with the Chinese Communist Party

Lastly, Wynn-Williams alleges that Facebook willingly cooperated with the demands of the Chinese Communist Party (CCP) in hopes of being allowed to operate in China. She says leadership was unbothered by the CCP’s authoritarianism, and she alleges that Zuckerberg lied to the US Congress about Facebook’s involvement with them.

(Shortform note: The US Congress’s concerns about Facebook’s involvement with the CCP may stem from a long history of tension between the US and China. In The Avoidable War, Kevin Rudd characterizes the US and China as opposites perpetually at odds with one another—the US being a democratic capitalist nation and China an authoritarian socialist nation. He explains that tensions trace back to the First Opium War (1839-1842), and he says that hostility between these nations has continued to escalate in the centuries since then. So by working with the CCP, Facebook allied itself with a longstanding rival of the US, raising red flags about its commitment to democratic principles and practices.)

According to Wynn-Williams, Facebook spent multiple years working with CCP officials in an attempt to gain access to the largest market left for it to reach. Zuckerberg championed this initiative, outlining via internal communications in 2014 his intent and a three-year plan. To him, it was the most important move Facebook could make to keep growing (which was his main goal).

(Shortform note: In The Divide, Jason Hickel describes economic reasoning that may explain why Zuckerberg felt so compelled to do business in China and cooperate with the CCP. In short, capitalism assumes that the economy needs to grow at a continuous exponential rate. As a result, modern businesses compete to grow faster and larger for longer. Zuckerberg was no exception: Despite the ethical dilemmas involved, he seemed to have felt that if Facebook wasn’t growing, it was dying. According to Hickel, the fallacy of this approach is that on a finite planet, infinite growth will eventually consume all resources. More to the point, nothing can grow forever—what goes up must come down.)

To operate in China, Facebook would have to abide by CCP regulations on media companies. These include giving the CCP the ability to censor speech on the platform, access user data and local servers, and broadly surveil citizens.

Through 2015 and 2016, Facebook worked with Chinese officials to develop technologies that would maintain what the CCP called “safe and secure social order” on the platform. Wynn-Williams, who wasn’t part of the China team until later on, recounts reading internal reports detailing Facebook’s engineering of special tools for moderation and censorship. They included mechanisms for keyword blocking, a viral post flagger, and a master switch that could wipe out viral content during politically sensitive times (like the anniversary of the Tiananmen Square incident).

(Shortform note: By 2018, China was running the world’s most advanced digital surveillance system. Their Great Firewall blocks foreign websites and employs algorithms to censor speech, and anyone who wants to register for social media has to use government ID. Meanwhile, a network of over 200 million cameras with facial recognition tracks citizens’ movements, and the police can access nearly any data, from health records to browsing history. The CCP ramped up development of these technologies through the 2010s—the same time period when Facebook was building censorship tech for them. Ironically, it’s possible that Facebook may have indirectly contributed to the same Great Firewall that now blocks the platform in China.)

Facebook also agreed to store Chinese users’ data in China—something it had refused to do for other governments like Russia or Brazil. According to Wynn-Williams, Facebook leaders knew that the CCP would access the servers and gather data. They also acknowledged that the Facebook employees building the censorship tools could be implicated in political violence if the CCP used them for ill.

In the end, Facebook didn’t succeed in gaining access to China. All the same, Wynn-Williams says, the episode demonstrated clearly that Facebook’s leaders were willing to ignore the ethical dilemma of working with an authoritarian regime if it meant they’d profit. In addition, Zuckerberg allegedly lied about it to the US Congress: When asked whether the platform would comply with CCP regulations, Zuckerberg said that “no decisions had been made.” This was four years after they’d begun working to enter China and did all of the above.

(Shortform note: In April 2025, Wynn-Williams testified before the US Congress about Facebook’s collaboration with the CCP, despite Meta’s attempts to silence her. As in Careless People, she alleged that Facebook built censorship tools including “virality counters” that flagged posts with over 10,000 views for review by a “chief editor.” She also said Facebook lied to “employees, shareholders, Congress, and the American public” about working with the CCP. Meta denies these allegations, stating they don’t currently operate services in China. However, Wynn-Williams claims their SEC filings show China is Meta’s second-biggest market (because Chinese companies, like Temu and Shein, spend billions on Facebook and Instagram ads.)

Averting a More Reckless Future

So far, we’ve laid out Wynn-Williams’ allegations that Facebook leaders were reckless, unaccountable, and ethically compromised in character and conduct; and that they caused major harm in the world. We’ll look next to her main takeaway.

Put simply, Wynn-Williams argues that Facebook’s recklessness continues to go unchecked—that Zuckerberg and company haven’t changed at all. She adds that this could be disastrous for the AI race between the US and China, in which both countries want to develop superior AI for military and economic use.

(Shortform note: Experts explain that the US-China AI race will be decided not by the best models, but by speed of adoption. AI will see application in the military, government, and private sectors, and the nation that’s first to spread and scale AI through these areas will likely come out on top.)

In 2024, Facebook’s leadership chose to open-source their AI models, or make them publicly available to license and build upon. In doing this, Wynn-Williams says, they’ve enabled Chinese tech firms (like DeepSeek) to compete with formerly dominant Western AI companies (like OpenAI) and potentially get an edge.

(Shortform note: In her 2025 testimony to Congress, Wynn-Williams noted how Meta’s open-sourcing of their Llama models allowed the Chinese firm DeepSeek to make major advances for the Chinese AI industry. This may be what she means when she says that Facebook leaders’ recklessness will continue to shape the world: They’ve directly contributed to the authoritarian side in what is potentially the world’s most consequential technological arms race.)

Why is Facebook’s recklessness a problem? Because if China wins the AI race, Wynn-Williams writes, the CCP will dominate the next generation of tech and use it for authoritarian ends rather than democratic ones. AI would allow them to go far beyond the tools Facebook was building for them and create tech for censorship, surveillance, and control with unprecedented power and precision.

(Shortform note: Authoritarian uses for AI might go even further than censorship and surveillance. China’s military has explicitly identified AI as “a strategic technology that will lead in the future,” and the PLA (People’s Liberation Army) is pursuing “algorithmic warfare” capabilities. These would include AI-enabled missiles with enhanced accuracy, autonomous underwater vehicles, and quantum computing applications for intelligence and surveillance. Some analysts warn that if China gains the upper hand in AI, it could use that influence to take a more aggressive approach toward Taiwan. In other words, AI may soon be impacting the geopolitical balances between major world powers.)

Finally, Wynn-Williams argues that if we want to ensure that the CCP doesn’t win the AI race, we need responsible leaders and stronger regulations that can keep people like Zuckerberg, Sandberg, and Kaplan from doing more harm.

(Shortform note: According to Roger McNamee, an early investor in Facebook and one-time adviser to Mark Zuckerberg, regulations won’t fix the problem if they target Facebook alone. He argues that all Big Tech companies rely on “surveillance capitalism,” an approach to business that involves collecting user data, using artificial intelligence to model and predict user behavior, and selling that information to advertisers. This is profitable because it lets advertisers target and drive buyers to their products more effectively than ever before. But it also violates people’s privacy and autonomy, according to McNamee. He says we need regulation to prevent harmful tech from making it to market, ban surveillance capitalism, and update antitrust laws.)