In public, Facebook seems to claim that it removes more than 90 percent of hate speech on its platform, but in private internal communications the company says the figure is only an atrocious 3 to 5 percent. Facebook wants us to believe that almost all hate speech is taken down, when in reality almost all of it remains on the platform.
This obscene hypocrisy was revealed amid the numerous complaints, based on thousands of pages of leaked internal documents, which Facebook employee-turned-whistleblower Frances Haugen and her legal team filed to the SEC earlier this month. While public attention on these leaks has focused on Instagram’s impact on teen health (which is hardly the smoking gun it’s been touted as) and on the News Feed algorithm’s role in amplifying misinformation (hardly a revelation), Facebook’s utter failure to limit hate speech and the simple deceptive trick it’s consistently relied on to hide this failure is shocking. It exposes just how much Facebook relies on AI for content moderation, just how ineffective that AI is, and the necessity to force Facebook to come clean.
In testimony to the US Senate in October 2020, Mark Zuckerberg pointed to the company’s transparency reports, which he said show that “we are proactively identifying, I think it’s about 94 percent of the hate speech we ended up taking down.” In testimony to the House a few months later, Zuckerberg similarly responded to questions about hate speech by citing a transparency report: “We also removed about 12 million pieces of content in Groups for violating our policies on hate speech, 87 percent of which we found proactively.” In nearly every quarterly transparency report, Facebook proclaims hate speech moderation percentages in the 80s and 90s like these. Yet a leaked a document from March 2021 says, “We may action as little as 3-5% of hate … on Facebook.”
Was Facebook really caught in an egregious lie? Yes and no. Technically, both numbers are correct—they just measure different things. The measure that really matters is the one Facebook has been hiding. The measure Facebook has been reporting publicly is irrelevant. It’s a bit like if every time a police officer pulled you over and asked how fast you were going, you always responded by ignoring the question and instead bragged about your car’s gas mileage.
There are two ways that hate speech can be flagged for review and possible removal. Users can report it manually, or AI algorithms can try to detect it automatically. Algorithmic detection is important not just because it’s more efficient, but also because it can be done proactively, before any users flag the hate speech.
The 94 percent number that Facebook has publicly touted is the “proactive rate,” the number of hate speech items taken down that Facebook’s AI detected proactively, divided by the total number of hate speech items taken down. Facebook probably wants you to think this number conveys how much hate speech is taken down before it has an opportunity to cause harm—but all it really measures is how big a role algorithms play in hate-speech detection on the platform.
What matters to society is the amount of hate speech that is not removed from the platform. The best way to capture this is the number of hate-speech takedowns divided by the total number of hate speech instances. This “takedown rate” measures how much hate speech on Facebook is actually taken down—and it’s the number that Facebook tried to keep secret.
Thanks to Haugen, we finally know the takedown rate, and it is dismal. According to internal documents, more than 95 percent of hate speech shared on Facebook stays on Facebook. Zuckerberg boasted to Congress that Facebook took down 12 million pieces of hate speech in Groups, but based on the leaked estimate, we now know that around 250 million pieces of hate speech were likely left up. This is staggering, and it shows how little progress has been made since the early days of unregulated internet forums—despite the extensive investments Facebook has made in AI content moderation over the years.
When the Iranian hacking group APT35 wants to know if one of its digital lures has gotten a bite, all it has to do is check Telegram. Whenever someone visits one of the copycat sites they’ve set up, a notification appears in a public channel on the messaging service, detailing the potential victim’s IP address, location, device, browser, and more. It’s not a push notification; it’s a phish notification.
Google’s Threat Analysis Group outlined the novel technique as part of a broader look at APT35, also known as Charming Kitten, a state-sponsored group that has spent the last several years trying to get high-value targets to click on the wrong link and cough up their credentials. And while APT35 isn’t the most successful or sophisticated threat on the international stage—this is the same group, after all, that accidentally leaked hours of videos of themselves hacking—their use of Telegram stands out as an innovative wrinkle that could pay dividends.
The group uses a variety of approaches to try to get people to visit their phishing pages in the first place. Google outlined a few scenarios it has observed lately: the compromise of a UK university website, a fake VPN app that briefly snuck into the Google Play Store, and phishing emails in which the hackers pretend to be organizers of real conferences, and attempt to entrap their marks through malicious PDFs, Dropbox links, websites, and more.
In the case of the university website, the hackers direct potential victims to the compromised page, which encourages them to log in with the service provider of their choice—everything from Gmail to Facebook to AOL is on offer—to view a webinar. If you enter your credentials, they go straight to APT35, which also asks for your two-factor authentication code. It’s a technique so old it’s got whiskers on it; APT35 has been running it since 2017 to target people in government, academia, national security, and more.
The fake VPN isn’t especially innovative, either, and Google says it booted the app from its store before anyone managed to download it. If anyone had fallen for the ruse, though—or does install it on another platform where it’s still available—the spyware can steal call logs, texts, location data, and contacts.
Frankly, APT35 are not exactly overachievers. While they convincingly impersonated officials from the Munich Security conference and Think-20 Italy in recent years, that too is straight out of Phishing 101. “This is a very prolific group that has a wide target set, but that wide target set is not representative of the level of success the actor has,” says Ajax Bash, security engineer at Google TAG. “Their success rate is actually very low.”
But the idea that Facebook’s technological underpinnings are supposedly so complex that they cannot be audited, and its business model so fast-moving that it cannot be slowed, is finally being overtaken by its undeniable dangers. A series of disastrous outcomes, from political manipulation of free elections to violence against minoritized populations to harm for young people, and even public health disinformation prolonging and worsening a pandemic, have destroyed the pleasant fiction that the company’s products produce a net positive for society.
In the case of the US auto industry, the need for not just regulation but also an enforcement agency to ensure compliance was similarly not immediately obvious. The necessity and wisdom of proactively regulating that infrastructural technology, instead of relying on the fiction that consumer choice was the primary mechanism for harm avoidance, was recognized only after decades of damage and years of whistleblowing and investigative journalism.
Like ubiquitous internet platforms, Detroit of the mid-20th century produced something most Americans felt they could not live without and quickly became dependent on. As suburban sprawl enveloped the areas around cities, racist resource allocation hollowed out urban centers and encouraged white flight. As a result, having one or more cars was increasingly a necessity to an ever larger number of Americans. State and federal government resources went toward creating ever more and wider roads to ensure that automobile traffic grew unchecked, even—or especially—at the expense of those who could not afford automobiles or who were structurally forbidden from moving out of neighborhoods increasingly cut in half and destroyed by eminent domain attempts to procure more land for the highways in and out of cities.
At this point in time, the Big Three also seemed unstoppable, rolling over the US landscape with the help of powerful business and government interests, while also illegally colluding with each other, and against public interest and public safety, for ever greater profits.
When the bombshell findings in lawyer and activist Ralph Nader’s 1965 bestseller Unsafe at Any Speed began to explode into US public discourse, auto executives lined up before Congress. They told the American public and those who represented them that they were doing their best to make cars safer and less polluting and that there was little they could do to immediately undo the harms produced by their product. Executives downplayed the scale of the public safety crisis and often claimed to be unaware of the extent of their products’ harms to consumers. Their answers were, of course, largely a charade aimed at saving profits and staving off regulation for as long as possible. The president of Ford at the time, Arjay Miller, recounted in vivid detail how his Lincoln Continental was safe enough to save his life when he got into a freeway accident—the doors didn’t jam, the gas tank didn’t explode, and Miller escaped unharmed. He pledged to ensure Ford did all they could in the coming years to improve safety even further.
But for years after that, Ford instead cut corners on safety, producing cars like the Ford Pinto that removed key safety features in order to get to market quickly and hold down manufacturing costs to reap maximum profit. In 1977, the infamous Ford Pinto “memo,” which was uncovered by Mother Jones investigative reporters, detailed the company’s horrifying cost analysis of past and future accidents. According to the memo, the gruesome deaths and full-body burns suffered by Pinto occupants in rear-end collisions amounted to an acceptable loss because, once lawsuits or other settlements were paid out, they would amount to less than the cost of fixing the Pinto design to prevent the gas tank from exploding. The cost of fixing the design was $11 per car. After public and governmental pressure, it was eventually implemented through a recall demanded by the recently created National Highway Traffic Safety Administration.
Today, a similar scenario is playing out in the realm of Big Tech—a term that has become shorthand for ad-driven platforms and internet-enabled arbitrage companies that lower the cost of goods and services by squeezing both workers and consumers. Whistleblowers from multiple companies, most of them women and many of them women of color, have stepped into the role Nader occupied in the ’60s—from Ifeoma Ozoma, who stood up to Pinterest and subsequently worked to create legislation to ban the abusive practice of nondisclosure agreements for whistleblowers in California, and Timnit Gebru, who alerted the world about Google’s lack of commitment to AI ethics in practice, to Sophie Zhang and now Frances Haugen. In each case, the companies have similarly attempted to silence, fire, or discredit these workers, reserving their harshest treatment for women of color.
The need for changing the power structures of this sector are critical not just for society but also for democracy: As Haugen’s testimony last week showed, Facebook marshaled its massive profits not toward fixing known problems but toward avoiding being perceived as having caused those problems. And just like Arjay Miller, Mark Zuckerberg has said whatever is needed to delay and deflect regulation.
Holiday shipping timelines will be tighter than ever this year. WIRED will be all over the Black Friday and Cyber Monday madness, but we’re also keeping our eyes peeled for early-bird discounts on tried-and-true gear, to ensure that your loved ones receive their gifts on time. This week, we found plenty of price drops on video games and gaming accessories.
Special offer for Gear readers: Get a1-year subscription toWIREDfor $5 ($25 off). This includes unlimited access to WIRED.com and our print magazine (if you’d like). Subscriptions help fund the work we do every day.
If you buy something using links in our stories, we may earn a commission. This helps support our journalism. Learn more.
Gaming Peripheral Deals
This is the lowest price ever on one of our favorite corded gaming headsets. It doesn’t look like it’s made for gamers, which might be a good thing, but it works with every console and PC. The soundstage is expansive, and the microphone was engineered with the help of Blue (the makers of the popular Blue Yeti USB mics).
This mouse got shuffled off our list of the Best Gaming Mice, but it’s still a solid pick. It’s very responsive, and the wireless design means there’s no cord to get in your way. There are also several buttons around it so you can create custom shortcuts and macros, plus Razer’s iconic RGB lighting makes the whole package aesthetically pleasing. This is one of the best discounts we’ve tracked, coming in at just $10 more than the lowest we’ve seen for the mouse. You can also snag the standard Razer Basilisk X Hyperspeed for $35 ($25 off)—it might be a better deal if you don’t care about tons of onboard buttons or the absolute lowest latency.
The Apex Pro lets you adjust the mechanical switches on each key, enabling faster response times and a more customized typing experience. It has tons of customizable color options as well, and there’s a tiny OLED display that shows system alerts, volume, and other settings you can tweak in SteelSeries’ software. This is just $5 more than the lowest price we’ve ever tracked.
The Razer Blade 15 is one of our favorite gaming laptops, and this edition gets you a few more bells and whistles packed into the same form factor. It features a 10th-gen Intel Core i7 processor, 512 GB of solid-state storage, 16 GB of RAM, and an NVIDIA GeForce RTX 2070 GPU. We’ve seen slightly better deals on this model, but this is still one of its lowest prices over the past several months.
This lightning-fast NVMe drive is great for nearly any gaming PC. It has super-fast read and write speeds, so you can expect speedy load times in your games. A few of us on the WIRED Gear team have been impressed by this drive (it’ll be added to our How to Build a PC guide soon!). For more storage, the 2-TB model is also discounted. Both are down to the lowest prices we’ve seen.
Video Game Deals
This deal is $2 more than one of the lowest prices we’ve seen for this game. It’s for a physical copy. Breath of the Wild isn’t the newest title, but it’s easily one of the Best Nintendo Switch Games around—it’s especially a must-try on the new Switch OLED.
This includes Final Fantasy Adventure, Secret of Mana, and Trials of Mana in one package. It’s a good gift for anyone with a fondness for the SNES console or these titles, but really, everyone should play these games at least once. The deal matches a low price we’ve seen a few other times.
This game has dipped to this price before, but it’s still a good deal. Kingdom Hearts III has great reviews across the net. Maybe you feel inspired after the recent announcement that Sora will be the final character in Super Smash Bros. Ultimate, or maybe you’re staring down winter hibernation with limited entertainment options. Either way, this is a great way to kill some time.
Street style meets storytelling in this recent release set in Shibuya, Tokyo. Your object is to solve a game within a game by exploring your environment and conquering negative emotions. Sounds like a good time to us.
Building your own PC is basically the opposite of free, but right now this simulator doesn’t cost a dime. It’s a great tool for figuring out your rig configuration ahead of actually assembling it. It uses real-world components to teach you how to build a desktop and troubleshoot potential hurdles. Pair it with PC Part Picker and our guide on How to Build a Desktop PC and you should feel prepared in no time.
For those unfamiliar, Humble Bundle offers a “pay what you want” structure featuring different curated game bundles. Part of the proceeds go to charity. This week’s bundle helps support the Breast Cancer Research Foundation and features a few different PC games that coincide with their board game counterparts. Chip in a few more bucks to get even more game downloads (like Carcassonne in the $9 bundle).
The unique gameplay mechanics and creepy setting make this one of the best games of 2019, but it still is an excellent title to try if you’ve never had the chance. The story has a surprising amount of depth, and did I say you get Jedi-like telekinesis powers?
Finally, we’ve reached bipartisan consensus on Big Tech, yay everyone! At least that’s the line the press is echoing ad nauseum. “Facebook Whistleblower Reignites Bipartisan Support for Curbing Big Tech,” the Financial Times trumpeted last week after Frances Haugen’s Senate testimony on Facebook. “Lawmakers Send Big Tech a Bipartisan Antitrust Message,” Newsweek wrote a day later. For more than a year, but especially after last week’s US Senate hearing, the media has been increasingly suggesting that Democrats and Republicans are setting aside their long-standing disagreements on tech policy.
But beyond their triumphant headlines, many of these articles note (often clumsily) that the “consensus” is merely an opinion that some kind of regulation for Big Tech is needed. This is where the idea of “bipartisan consensus” crumbles, and where the danger in this expression lies.
It’s true that in the past few years American lawmakers have become far more outspoken about Silicon Valley technology giants, their products and services, and their market practices. Yet merely agreeing that something must be done, and on that alone, is about as superficial as bipartisan consensus gets. Elected representatives of both parties still have disagreements about what that something is, why that something should happen, and what the problems are in the first place. All these factors are shaping both the proposed regulations in Congress and the path forward to making them reality.
On top of this, the media separating national politics from the tech legislation process only threatens to repeat the problems of the last several decades, where imagining technology as non-political is conducive to regulators and society ignoring dangers right in front of them. This overblown rhetoric skews analysis of the difficult road ahead to real, substantive regulation—and just how many threats to democracy (and democratic tech legislation) lie from within.
For decades, liberal democracies from the United States to France to Australia consistently touted the internet as a free, secure, and resilient golden child of democracy. US leaders in particular, from Bill Clinton’s Jell-O-to-a-wall speech in 2000 to the State Department’s so-called internet freedom agenda of 2010, hailed the web’s power to topple authoritarianism worldwide. Left alone, this logic went, democratic governments could enable the internet to be as pro-democracy as possible.
The groundswell of calls to regulate Big Tech today is no small change. While it’s tempting to see this shift as unilateral, some parts of the media often forget that tech is not a monolith and that many disparate incidents have led to many disparate calls for regulation: the Equifax data breach, the Cambridge Analytica privacy scandal, Russian ransomware attacks, Covid misinformation, disinformation campaigns targeting Black voters, uses of racist and sexist algorithms, abusive police uses of surveillance technology, and on and on. Not all lawmakers care equally, or at all, about these issues.
Data breaches and ransomware would seem to be two areas with the greatest potential for consensus legislation; members of Congress hardly stand up to profess their belief in lowering the cybersecurity bar and making their constituents vulnerable to attacks. Earlier this year, after multiple, significantly damaging ransomware attacks launched from within Russia, members of both parties condemned the behavior and highlighted how Congress and the White House could respond by sanctioning Russian actors and investing more in domestic security. The House and Senate held ransomware hearings in July, building on important civil society work to drive bipartisan responses to the threat.