Translation: We fired and harassed out our top AI researchers from historically marginalized groups who did the work for us, and we are now looking for more people from historically marginalized groups to burn out, exploit, and expend.
I encourage students from historically marginalized groups who are interested in learning to conduct research in AI/ML, CS or related areas to consider applying for our CSRMP mentorship program! We have 100s of researchers @GoogleAI who are excited to work with you. twitter.com/GoogleAI/statu…
As regulators seek ways to curb the company’s power, there is more focus on the vast index — hundreds of billions of web pages — behind its search engine.
In 2000, just two years after it was founded, Google reached a milestone that would lay the foundation for its dominance over the next 20 years: It became the world’s largest search engine, with an index of more than one billion web pages.
The rest of the internet never caught up, and Google’s index just kept on getting bigger. Today, it’s somewhere between 500 billion and 600 billion web pages, according to estimates.
Now, as regulators around the world examine ways to curb Google’s power, including a search monopoly case expected from state attorneys general as early as this week and the antitrust lawsuit the Justice Department filed in October, they are wrestling with a company whose sheer size has allowed it to squash competitors. And those competitors are pointing investigators toward that enormous index, the gravitational center of the company.
“If people are on a search engine with a smaller index, they’re not always going to get the results they want. And then they go to Google and stay at Google,” said Matt Wells, who started Gigablast, a search engine with an index of around five billion web pages, about 20 years ago. “A little guy like me can’t compete.”
Understanding how Google’s search works is a key to figuring out why so many companies find it nearly impossible to compete and, in fact, go out of their way to cater to its needs.
Every search request provides Google with more data to make its search algorithm smarter. Google has performed so many more searches than any other search engine that it has established a huge advantage over rivals in understanding what consumers are looking for. That lead only continues to widen, since Google has a market share of about 90 percent.
Google directs billions of users to locations across the internet, and websites, hungry for that traffic, create a different set of rules for the company. Websites often provide greater and more frequent access to Google’s so-called web crawlers — computers that automatically scour the internet and scan web pages — allowing the company to offer a more extensive and up-to-date index of what is available on the internet.
When he was working at the music site Bandcamp, Zack Maril, a software engineer, became concerned about how Google’s dominance had made it so essential to websites.
In 2018, when Google said its crawler, Googlebot, was having trouble with one of Bandcamp’s pages, Mr. Maril made fixing the problem a priority because Google was critical to the site’s traffic. When other crawlers encountered problems, Bandcamp would usually block them.
Mr. Maril continued to research the different ways that websites opened doors for Google and closed them for others. Last year, he sent a 20-page report, “Understanding Google,” to a House antitrust subcommittee and then met with investigators to explain why other companies could not recreate Google’s index.
“It’s largely an unchecked source of power for its monopoly,” said Mr. Maril, 29, who works at another technology company that does not compete directly with Google. He asked that The New York Times not identify his employer since he was not speaking for it.
A report this year by the House subcommittee cited Mr. Maril’s research on Google’s efforts to create a real-time map of the internet and how this had “locked in its dominance.” While the Justice Department is looking to unwind Google’s business deals that put its search engine front and center on billions of smartphones and computers, Mr. Maril is urging the government to intervene and regulate Google’s index. A Google spokeswoman declined to comment.
Websites and search engines are symbiotic. Websites rely on search engines for traffic, while search engines need access to crawl the sites to provide relevant results for users. But each crawler puts a strain on a website’s resources in server and bandwidth costs, and some aggressive crawlers resemble security risks that can take down a site.
Since having their pages crawled costs money, websites have an incentive to let it be done only by search engines that direct enough traffic to them. In the current world of search, that leaves Google and — in some cases — Microsoft’s Bing.
Google and Microsoft are the only search engines that spend hundreds of millions of dollars annually to maintain a real-time map of the English-language internet. That’s in addition to the billions they’ve spent over the years to build out their indexes, according to a report this summer from Britain’s Competition and Markets Authority.
Google holds a significant leg up on Microsoft in more than market share. British competition authorities said Google’s index included about 500 billion to 600 billion web pages, compared with 100 billion to 200 billion for Microsoft.
Other large tech companies deploy crawlers for other purposes. Facebook has a crawler for links that appear on its site or services. Amazon says its crawler helps improve its voice-based assistant, Alexa. Apple has its own crawler, Applebot, which has fueled speculation that it might be looking to build its own search engine.
But indexing has always been a challenge for companies without deep pockets.
The privacy-minded search engine DuckDuckGo decided to stop crawling the entire web more than a decade ago and now syndicates results from Microsoft. It still crawls sites like Wikipedia to provide results for answer boxes that appear in its results, but maintaining its own index does not usually make financial sense for the company.
“It costs more money than we can afford,” said Gabriel Weinberg, chief executive of DuckDuckGo. In a written statement for the House antitrust subcommittee last year, the company said that “an aspiring search engine start-up today (and in the foreseeable future) cannot avoid the need” to turn to Microsoft or Google for its search results.
When FindX started to develop an alternative to Google in 2015, the Danish company set out to create its own index and offered a build-your-own algorithm to provide individualized results.
FindX quickly ran into problems. Large website operators, such as Yelp and LinkedIn, did not allow the fledgling search engine to crawl their sites. Because of a bug in its code, FindX’s computers that crawled the internet were flagged as a security risk and blocked by a group of the internet’s largest infrastructure providers. What pages they did collect were frequently spam or malicious web pages.
“If you have to do the indexing, that’s the hardest thing to do,” said Brian Schildt Laursen, one of the founders of FindX, which shut down in 2018.
Mr. Schildt Laursen launched a new search engine last year, Givero, which offered users the option to donate a portion of the company’s revenue to charitable causes. When he started Givero, he syndicated search results from Microsoft.
Most large websites are judicious about who can crawl their pages. In general, Google and Microsoft get more access because they have more users, while smaller search engines have to ask for permission.
“You need the traffic to convince the websites to allow you to copy and crawl, but you also need the content to grow your index and pull up your traffic,” said Marc Al-Hames, a co-chief executive of Cliqz, a German search engine that closed this year after seven years of operation. “It’s a chicken-and-egg problem.”
In Europe, a group called the Open Search Foundation has proposed a plan to create a common internet index that can underpin many European search engines. It’s essential to have a diversity of options for search results, said Stefan Voigt, the group’s chairman and founder, because it is not good for only a handful of companies to determine what links people are shown and not shown.
“We just can’t leave this to one or two companies,” Mr. Voigt said.
When Mr. Maril started researching how sites treated Google’s crawler, he downloaded 17 million so-called robots.txt files — essentially rules of the road posted by nearly every website laying out where crawlers can go — and found many examples where Google had greater access than competitors.
ScienceDirect, a site for peer-reviewed papers, permits only Google’s crawler to have access to links containing PDF documents. Only Google’s computers get access to listings on PBS Kids. On Alibaba.com, the U.S. site of the Chinese e-commerce giant Alibaba, only Google’s crawler is given access to pages that list products.
This year, Mr. Maril started an organization, the Knuckleheads’ Club (“because only a knucklehead would take on Google”), and a website to raise awareness about Google’s web-crawling monopoly.
“Google has all this power in society,” Mr. Maril said. “But I think there should be democratic — small d — control of that power.”
New feature can deep-link to specific text on a Web page, with highlighting.
Google has been cooking up an extension to the URL standard called “Text Fragments.” The new link style will allow you to link not just to a page but to specific text on a page, which will get scrolled to and highlighted automatically once the page loads. It’s like an anchor link, but with highlighting and creatable by anyone.
The feature has actually been supported in Chrome since version 80, which hit the stable channel in February. Now a new extension from Google makes it easy to create this new link type, which will work for anyone else using Chrome on desktop OSes and Android. Google has proposed the idea to the W3C and hopes other browsers will adopt it, but even if they don’t, the links are backward-compatible.
The syntax for this URL is pretty strange looking. After the URL, the magic is in the string “#:~:text=” and then whatever text you want to match. So a full link would look like this:
If you copy and paste this into Chrome, the browser will open Wikipedia’s cat page, scroll to the first text that matches “Most breeds of cat have a noted fondness for sitting in high places,” and will highlight it. If the text doesn’t match anything, the page will still load. Backward-compatibility works because browsers currently support the number sign (#) as a URI fragment, which usually gets used for anchor links that are made by the page creator. If you paste this into a browser that doesn’t support it, the page will still load, and everything after the number sign will just be ignored as a bad anchor link. So far, so good.
One problem is that this means you can have spaces in a URL. On a webpage or forum, you can hand-code the link with a href tag (or whatever the non-HTML equivalent is) and everything will work. For instant messengers and social media though, which don’t allow code and use automatic URL parsers, things get a bit more complicated. Every URL parser treats a space as the end of a URL, so you’ll need to use percent-encoding to replace all the spaces with the equivalent “%20.” URL parsers now have a shot at linkifying this correctly, but it looks like a mess:
Spaces aren’t the only characters that can cause problems. The standard RFC 3986 defines several “reserved” characters as having a special meaning in a URL, so they shouldn’t be in a URL. Web-page-authoring tools tend to handle these characters automatically, but now that you’re embedding arbitrary sentences in a URL for highlighting, there’s a higher chance you’ll run into one of these reserved characters:! * ‘ ( ) ; : @ & = + $ , / ? # [ ]. They all need to be percent-encoded in order for the URL to work, and Google’s extension takes care of that for you.
Google’s new Chrome extension, called “Link to Text Fragment,” (it’s also on Github) will put a new entry in Chrome’s right-click menu. You just highlight text on a page, right-click it, and hit “Copy link to selected text.” Like magic, a text fragment link will end up on your clipboard. All the text encoding is done automatically, so the link should work with most websites and messengers.
Google seems like it is going to start pushing out support for text fragments across its Web ecosystem, even without the W3C. The links have already started to show up in some Google search results, which allow Chrome users to zip right to the relevant text. It’s probably only a matter of time before link creation moves from an extension to a normal Chrome feature.
Some of its employees tried to stop their company from doing work they saw as unethical. It blew up in their faces.
Laurence Berland had just gotten out of the subway in New York, some 3,000 miles from his desk in San Francisco, when he learned that Google had fired him. It was the Monday before Thanksgiving, and the news came to him, bad-breakup-style, via email. “Following a thorough investigation, the company has found that you committed several acts in violation of Google’s policies,” the note said. It did not elaborate on what he had done to violate these policies.
Berland, an engineer who had spent more than a decade at the company, had reason to expect he might be fired. He had been suspended a few weeks earlier after subscribing to the open calendars of several senior Google employees, whom he suspected of meeting with outside consultants to suppress organizing activity at the company. During a subsequent meeting at which he was questioned by Google investigators, he had the feeling that they were pressuring him to say something that could be grounds for termination. Then, the Friday before he was fired, he had spoken at a well-publicized rally of his co-workers outside Google’s San Francisco offices, accusing the company of silencing dissent.
Even so, the timing and manner of his dismissal surprised him. “I thought they’d do it when all the media attention died down,” he said. “When the suspensions and the rally were no longer on people’s minds.” Instead, at a moment when the spotlight was shining brightly, Google had escalated — as if to make a point.
Berland was one of at least four employees Google fired that day. All four were locked in an ongoing conflict with the company, as they and other activists had stepped forward to denounce both its treatment of workers and its relationship with certain customers, like U.S. Customs and Border Protection.
Berland’s terminated colleagues were even more shocked by the turn of events than he was. Rebecca Rivers, a software engineer based in Boulder, Colo., was dismissed over the phone after accessing internal documents. Rivers had only recently come out as transgender and was pursuing a medical transition. “I came out at Google expecting to stay at Google through the entire transition,” she said. “It’s terrifying to think about going to a job interview, because I’m so scared of how other companies treat trans employees.”
Sophie Waldman and Paul Duke, the two other Googlers fired that day, had not received so much as a warning, much less a suspension. Though they had been questioned by corporate security two months earlier about whether they had circulated documents referring to Customs and Border Protection contracts, they had been allowed to continue their work without incident. Waldman, a software developer in Cambridge, Mass., said she was given a 15-minute notice before she was summoned to the meeting where she was fired; Duke, an engineer in New York, said an invitation appeared on his calendar precisely one minute beforehand. Security officials escorted him out of the building without letting him return to his desk. “I had to describe to them what my jacket, scarf and bag looked like,” he said.
From its earliest days, Google urged employees to “act like owners” and pipe up in all manner of forums, from mailing lists to its meme generator to open-ended question-and-answer sessions with top executives, known as T.G.I.F. It was part of what it meant to be “Googley,” one of the company’s most common compliments. So well entrenched was this ethic of welcoming dissent that the company seemed to abide by it even after the uprising began, taking pains to show it was heeding activists’ concerns.
Over the past year, however, Google has appeared to clamp down. It has gradually scaled back opportunities for employees to grill their bosses and imposed a set of workplace guidelines that forbid “a raging debate over politics or the latest news story.” It hastried to prevent workers from discussing their labor rights with outsiders at a Google facility and even hired a consulting firm that specializes in blocking unions. Then, in November, came the firing of the four activists. The escalation sent tremors through the Google campus in Mountain View, Calif., and its offices in cities like New York and Seattle, prompting many employees — whether or not they had openly supported the activists — to wonder if the company’s culture of friendly debate was now gone for good.
(A Google spokeswoman would not confirm the names of the people fired on Nov. 25. “We dismissed four individuals who were engaged in intentional and often repeated violations of our longstanding data-security policies,” the spokeswoman said. “No one has been dismissed for raising concerns or debating the company’s activities.” Without naming Berland, Google disputed that investigators pressured him.)
As similar forms of worker activism have spread to other companies, including Amazon and Microsoft, it has also raised deeper questions about the nature of the entire industry. Silicon Valley has often held itself up as a highly evolved ecosystem that defies the usual capital-labor dichotomy — a place where investors, founders, executives and workers are all far too dependent on one another to make anything so crass as class warfare. The recent developments at Google have thrown that egalitarian story into doubt, showing that even in the most rarefied corners of Silicon Valley, the bosses are willing to close ranks and shut down debate when the stakes are high enough. The fate of these activists, meanwhile, forces America’s white-collar professionals to grapple with an uncomfortable thought: If the nation’s most sought-after workers can’t stop their employer from behaving in ways that they deplore, where does that leave the rest of us?
As recently as a decade ago, the prospect of labor unrest at Google, of all workplaces, would have seemed laughable. From the very beginning, its founders embraced the view that the value of its business hinged far more on the brainpower of its workers than on any particular lines of code it might own. “Google is organized around the ability to attract and leverage the talent of exceptional technologists and businesspeople,” Larry Page, a founder, wrote in the company’s 2004 I.P.O. prospectus. “We will reward and treat them well.”
To this end, the founders not only made Google one of the cushiest workplaces in corporate America, with its screening rooms, nap pods and mounds of free snacks. They also made it one of the most idealistic. “Talented people are attracted to Google because we empower them to change the world,” Page added in his letter. He was channeling a worldview with roots in the 1960s, when the engineers in Northern California’s nascent computer industry imbibed the local consciousness-raising counterculture and emerged with a faith in the power of technology to save humankind. As so much of the world economy moved online and the company’s profits soared, Google and the rest of Silicon Valley developed a reputation as the American dream factory, the place where the world’s smartest young people wanted to be — where technologists and businesspeople could pull down staggering incomes while still believing that they were engaged in an act of altruism.
But as the world they were remaking came into focus, many of those engineers had second thoughts. With social media swallowing up the public square, it was hard not to notice that Google and Facebook had become an advertising duopoly with an unsettling grip on the entire world’s attention. And after the 2016 presidential election, the consequences of the social media revolution came to seem dystopian. Many engineers felt deeply anguished at the news that foreign governments had exploited their technology in an attempt to influence domestic politics. “It showed that pretty much any system large enough and complex enough can be co-opted for nefarious purposes,” Rivers said.
Others became increasingly concerned that the Trump administration might now use their tools in service of policies they found immoral. Unknown to them, Google was at work on a project that would bring these anxieties right to their own work spaces. In September 2017, the company quietly entered into a contract to help the U.S. Department of Defense track people and vehicles in video footage captured by drones. During a meeting that December, Google presented initial results that showed its artifical-intelligence software was more successful than human data labelers at identifying vehicles, according to an internal Google document we reviewed. By February, the effort, known as Project Maven, was slotted into a launch calendar for soon-to-be-deployed products.
Google’s involvement in the project remained under wraps within the company for months. But by February, rumors about it had seeped out beyond the small circle of engineers who were doing the work. Concerned employees began to search through code and documents and compile their findings. One activist, an engineer named Liz Fong-Jones, called attention to Project Maven in an internal blog post, according to other workers who saw it, and the circle of concern grew. Many of the engineers feared that the technology would be used to single out targets for killing. A growing number of workers concluded that the program was not in keeping with Google’s values, and they wanted management to know about it. The discovery that their employer was working hand in hand with the Defense Department to bolster drone warfare “was a watershed moment,” said Meredith Whittaker, an A.I. researcher based in New York who led Google’s Open Research Group. “If they were able to do that without any internal backlash, dissent, we would have crossed a significant line.”
Whittaker, along with some colleagues, drafted a letter petitioning management to cancel the program, which quickly amassed thousands of signatures after they published it internally. In April, company leaders also flew Whittaker to Google’s headquarters in Mountain View for three back-to-back town-hall discussions about the A.I. project, which it broadcast to Google employees around the world.
The discussions did not go well for the company. Whittaker’s concerns that the technology would enable extrajudicial killings were met with platitudes, or worse. One participant, according to Whittaker and workers who watched, compared Google’s technology to a hammer: something that could be used to harm or to build. Diane Greene, then the chief executive of Google’s cloud business, said the government was essentially using the same off-the-shelf software any customer could buy, a claim that the activists believed to be false. Even Googlers sympathetic to the contract found the presentation lacking. Many workers cheered Whittaker on in a series of memes, including one — echoing a popular meme about Senator Elizabeth Warren of Massachusetts — with the words “nevertheless she persisted” superimposed on her photo.
And her performance seemed to yield results. At an internal town hall in June 2018, Greene said Google would not seek to renew the contract, which internal emails showed could have been worth up to $250 million per year. (The Defense Department declined to comment on Project Maven’s vendors or its military applications.) The about-face shocked many Google employees, even as it gave them a sense of their growing power. Greene soon stepped aside from her role leading the cloud business.
Galvanized, the activists began to press further. They quickly honed a formula for pushing back on products they opposed:
research the project using Google’s internal search tools,
compile the findings into a Google Doc,
write a protest letter and gather co-worker signatures,
debate the issue on Google Plus.
Over the summer, another secret program, nicknamed Dragonfly, came to light in The Intercept. The project would censor search results in China on behalf of the Chinese government, and after months of internal protest, Google appeared to back away from that program too. During congressional testimony that December, the company’s chief executive, Sundar Pichai, said he had “no plans to launch a search service in China,” though some activists worried it might yet return.
Workers weren’t just organizing to save the world from Google. They were also organizing to save themselves from Google, where those who didn’t fit the mold of the straight, white, male techie felt they could be too easily marginalized or dismissed.
The previous year, after a Google software engineer named James Damore circulated an essay he had written arguing that women were less suited biologically for careers in technology than men, a handful of Google employees pushed back with a memo of their own. “We must remember that we are ultimately all affected by technology,” they wrote, “and that every one of us should have a voice in how it’s built.”
The eight authors of the memo were part of a growing group of workers who spent the next 15 months laying the groundwork for a wider counteroffensive. They knew that unionizing would be a tall order in an industry in which workers had long regarded unions suspiciously. They settled on a more modest plan: to quietly confer with colleagues about the workplace issues they most wanted to address.
By 2018, the moment was ripe. That October, The New York Times revealed that the company had given Andy Rubin, a former executive, a $90 million exit package after an employee’s sexual-misconduct complaint against him was deemed credible. (Rubin denied any wrongdoing and blamed a smear campaign for the allegations.) Amid the outrage that followed the revelation, Claire Stapleton, a marketing manager for YouTube, which Google bought in 2006, suggested that women walk out. Stapleton, who had not been involved in previous activism, thought the walkout would demonstrate that women are pivotal to Google’s work force.
A more seasoned activist, worried that not enough people would walk out, suggested that Stapleton give colleagues other ways of participating, like putting an out-of-office message on their inboxes. But Stapleton forged ahead, figuring the walkout would work if even a few hundred people showed up. The organizers took to their mailing lists to get the word out.
The idea caught on swiftly, though, and on Nov. 1, some 20,000 Google employees around the world left their offices to protest the payout and other workplace frustrations. Critically for a tech company, it wasn’t just well-paid engineers and product leads who turned up, but also clerical and maintenance workers. Many were contractors and temps determined to challenge their second-class status.
Managers took notice. During a subsequent staff meeting, Google’s vice president for human resources, Eileen Naughton, said the Rubin revelations had created a mess for the company. “We’re cleaning up after the circus elephants here,” she said, going on to note that Rubin had left the company and that no payouts had been awarded to harassers since she had stepped into her role.
Activists capitalized on the success. In San Francisco and New York, they coordinated weekly “walkout lunches” to keep their co-workers involved and mobilize against the next workplace outrage, whether it was a product that threatened humanity or a human-resources practice that threatened their livelihood. The more they talked, the more they realized their concerns had a common thread: Executives had too much power over the company, and they had too little. They wanted more. They organized chat groups on encrypted apps like Signal, with innocuous names like “care package delivery” so that they wouldn’t be outed if a manager glimpsed their phones. They prepared tip sheets to help workers approach colleagues, in hopes of building a permanent organization.
Some senior Google executives spoke approvingly about the walkout, but the company also made clear there were limits to its tolerance for worker protests.
The approach had a rich pedigree in the tech sector. In 1990, Apple announced changes to a profit-sharing plan that sent employees to the company’s internal message boards to vent their outrage. Apple soon backtracked. But when about 50 workers formed an organization called Employees for One Apple, which demanded more influence over company decisions, Apple’s leadership granted only token concessions. The organization proved largely powerless, according to a paper by the researchers Libby Bishop and David Levine.
Alan Hyde, an expert on labor law at Rutgers Law School in Newark who has written extensively about the tech sector, said the pattern has repeated itself again and again in Silicon Valley: Management missteps, irate workers take to their mailing lists and message boards and leaders sometimes back down. “Then it dissipates,” he said. “No permanent organizations form, no organization that can plan and act strategically.”
Google seemed to be following a similar playbook. Activists workshopped the questions they would ask at staff meetings to force executives to explain why they weren’t responding to their walkout demands, which included a request for an employee representative on the board and an end to forced arbitration for all claims by employees.
Google would soon fold on arbitration. But it also announced that it would no longer take in-person questions at companywide meetings. Hereafter, all questions would have to be submitted through Dory, the company’s digital question platform.
The Case Against GoogleFeb. 20, 2018
The workers continued to agitate, with mixed results. In April 2019, Kent Walker, Google’s senior vice president for global affairs and top lawyer, convened an external council to advise the company on its use of artificial intelligence. But workplace activists discovered that one appointee, the president of the Heritage Foundation, Kay Cole James, had made what some employees saw as anti-L.G.B.T.Q. and anti-immigrant comments. This generated a wave of embarrassing publicity and, just over one week after he announced the council, Walker pulled the plug.
Soon the most troublesome activists began to disappear from the company. Whittaker, who had spoken out against James’s appointment, had been close to finding a new home within the company after being told that her specialty, ethical questions relating to artificial intelligence, was no longer a priority for Google Cloud. Shortly after the council was disbanded, an official in the new group told her the transfer was dead. Her prospects for remaining at the company appeared to dim.
Stapleton, for her part, said that managers took responsibilities and direct reports away from her and that an executive suggested she go on medical leave until the dispute blew over, even though she had no illness. “When people are speaking up — and they are because that’s the Google way — it’s not good for your career,” she said. Both women soon left the company. The company said that it changed the structure of Stapleton’s team as part of a routine shift but did not demote her, and that Whittaker was asked to return to an earlier role and refused to do so. Whittaker said the specific role was not one she had held before and was effectively a demotion.
It was around this time that Google appeared to be preparing for a more aggressive approach to confronting activists. In May, Walker sent out a companywide email informing workers that they could be disciplined or fired for accessing material that was supposed to be viewed by only those with a “need to know.” The email styled itself as a simple reminder about policy changes the company had recently made. (Subject line: “An important reminder on data classifications.”)
Many Googlers accepted the change as a natural shift for a company with increasingly high-profile and publicity-averse customers. A Google spokeswoman said that the underlying policy had been in place since 2007 but that the language is periodically updated to make it easier to understand and apply. And admirers and critics alike say Walker, who served as an assistant U.S. attorney before migrating to Big Tech, is evangelical about Google’s power to do good, and that he has vastly expanded his reach over the rest of the organization.
But other accounts put Walker’s vision in a different light. Ross LaJeunesse, a top public-policy official at the company, had long been concerned that Google’s cloud business was drawing the company into a web of relationships with repressive foreign governments and other questionable actors. “It makes us an accessory if we are hosting their email systems or their data,” he said in an interview. LaJeunesse, who left the company last year and is currently running for the U.S. Senate in Maine, took his concerns to Walker in 2017 and suggested creating a human rights program that would vet deals. But LaJeunesse said Walker repeatedly demurred, eventually responding that such a program would itself expose Google to legal liability. (Google denied the account. “Google’s commitment to human rights is unwavering,” the spokeswoman said. “While Ross had lobbied to have his group oversee human rights issues across our products, we felt it was more effective to have product and functional teams directly handle these issues.”)
Now activists were particularly troubled by the vagueness of Walker’s dictum: Sensitive material would not necessarily be labeled “need to know.” The onus would be on workers to determine whether they should look at it or not. “In my orientation, I was encouraged to read all the design documents I could find, look at anything about how decisions are made,” said Duke, the New York engineer. “Now they’re saying that’s no longer OK. That is a major shift in culture.” (The Google spokeswoman denies that a shift has occurred. “Google has a rich history of employees’ raising concerns and debating company decisions, but flagrant violation of our policy has never been part of our culture,” she said.)
To some activists, the policy shift appeared to be a recipe for a kind of digital entrapment, or worse. “I said at the time that they intended to fire whoever finds the next Maven,” said Irene Knapp, a Google engineer and activist who left the company in September.
For six months after the 2018 walkout, Google’s founders, Larry Page and Sergey Brin, had little to say to their workers. There was no mass email with their names on it, no sighting of them at gatherings of rank-and-file employees. Googlers were beginning to take notice. One popular internal meme, playing on the error message a web browser displays when a site won’t load, joked, “404 Page not found.”
The founders suddenly materialized at a companywide meeting in May 2019. But it wasn’t to reassert control over the company. According to multiple employees who tuned in, the two men made idle small talk with Pichai, the chief executive, about what they had been up to — Page said he’d been working with Sidewalk Labs, a subsidiary of Google’s parent company, Alphabet, that focused on urban problems like housing; Brin said he’d been working with Wing, the company’s delivery-drone subsidiary. They mostly ducked employees’ questions, including one about whether the company had retaliated against Stapleton.
The main presentation was from Thomas Kurian, the new head of the company’s cloud division. Kurian spoke about how he planned to increase Google Cloud’s revenue by competing with Microsoft and Amazon. The two rivals had a significant head start, but Kurian promised to catch up, rapidly. It was a message that Kurian, a longtime Oracle executive, had been pounding internally since he arrived at the company earlier in the year.
“He said we are a distant third in the cloud market,” an engineer in the cloud department recalled, echoing other sources. “If we don’t make it to the top two, maybe we should not be in this market anymore.”
Kurian focused on hiring more sales and customer-support personnel and made clear that he was eager to do business with the government. At one point, Whittaker recalled her manager’s telling her that Kurian aspired to be “everywhere Lockheed is.” (Google said Kurian never made a direct comparison between Google’s business and Lockheed Martin’s defense work.) And in July, Kurian got an opportunity: Customs and Border Protection announced the first step toward bidding out a major information-technology contract.
The idea of working with C.B.P. filled many Googlers with dread, and they began to circulate a petition urging executives to spurn the agency. “The winning cloud provider will be streamlining C.B.P.’s infrastructure and facilitating its human rights abuses,” the petition said. “It’s time to stand together again and state clearly that we will not work on any such contract.” (The agency declined to comment for this article.)
Rebecca Rivers quickly signed, then began to wonder: How deeply was Google already entangled with the agency? She turned to Moma, a company intranet that employees have used to find everything from cafeteria menus to design documents, and quickly unearthed emails referencing several projects sold through third-party vendors, including one contract for about $15,000 worth of Google Cloud services, a roughly $250,000 contract for a version of Google Chrome and a third contract for about $600,000 with Google Maps.
Rivers had personally built features for Maps, including one that flagged recent changes to locations of interest. She thought it might have humanitarian uses, like locating refugees fleeing war or climate catastrophe, or could just help retailers pick the sites of new stores. “The worst I thought it could be used for was making rich people more rich,” she said. But it could be used by anyone, she later realized, including an agency that separated children from their parents. “It was gut-wrenching,” she said. “It was the software I wrote that I was most proud of.” Rivers said she passed the information to Sophie Waldman, a co-worker, who combined it with information she had found and disseminated it to other colleagues. Before long, many realized that they, too, had been unwittingly working on technology that could benefit the agency.
The biggest uproar surrounded a project called Anthos, a program that allows customers to combine their existing cloud services and Google’s. According to a report in Business Insider, internal documents showed that Google had given C.B.P. a trial of Anthos. Some engineers working on the project were enraged. They had been told that Anthos was intended for banks and other businesses. Workers took to internal mailing lists to express their outrage about the project.
Kurian had previously said the cloud department had nothing to do with the “Southern border,” according to employees. Now he reiterated the point on one of the threads, adding that in any case, Anthos was just a trial.
Many of the engineers were satisfied with Kurian’s response. “I wish we didn’t do those things,” the engineer in the cloud department said. “If someone is going to do them, I don’t trust Microsoft. At least we could have some oversight.”
Google officials insist that the decision to consider working with C.B.P. was not a statement about the company’s values. “Every experience I’ve had past or present, Google is a deeply values-driven, deeply mission-driven, deeply principled company,” said Jennifer Fitzpatrick, a senior vice president at the company. “But these are not areas where consensus is possible or even likely.” She added: “We do have to make decisions. We do have to move forward.”
But to the activists, Kurian’s explanation was nonsensical. There didn’t seem to be anything in the contracts that restricted Google from contributing to work on the Southern border. The I.T. infrastructure Google was providing would support the agency’s operations everywhere. And regardless of the merits of the contracts, there was the disturbing fact that the engineers working on them had been misled about the purpose of the technology they were creating. Some had decided to work on Anthos precisely because it did not appear to be destined for the national-security apparatus. “If workers aren’t told what the real purpose of their work is, they have no agency in deciding whether or not they want to help with those things,” Berland said. “They become unwittingly complicit.” (Google declined to comment specifically on Anthos.)
By last fall, workers and management appeared to be locked in a game of brinkmanship. The company tried to cancel a meeting about unions and labor rights that its Zurich employees had arranged with two union officials, but the employees disregarded messages from office leadership and held the meeting anyway.
Around the same time, employees discovered that Google officials had been meeting with a firm called IRI Consultants since at least May. The firm has done work helping to defeat organizing campaigns at hospitals and other workplaces, in one case by instructing managers to play up the history of Mafia influence on organized labor. (IRI did not return calls seeking comment.) Google also brought on a former assistant U.S. attorney with expertise prosecuting white-collar crime, Stephen C. King, as a top internal security official, in addition to a former Colorado police officer it hired earlier in the year.
It was during this time that the practical effect of the need-to-know policy began coming into focus. In September, the internal security team interviewed a handful of employees who had been involved in circulating the petition asking Google not to work with Customs and Border Protection and in unearthing the documents showing that Google already was.
According to Rivers, who was among those interviewed, one investigator asked how she found the documents and whom she shared them with, a potentially legitimate line of questioning in light of the new policy. But, she said, he seemed more interested in what she knew about how the petition was planned. “It was about how we’re organizing,” Rivers recalled. “How many people there were, the methods we used to meet.” (Google said that policy violations were the focus of the interview and that it appropriately conducted the investigations, giving individuals a chance to explain their actions.)
Rivers and the other employees interviewed believed that the questions about the documents were a pretext for trying to restrict their organizing. She and some of the others considered filing a charge with the National Labor Relations Board, but they worried that this would put them further at risk. “We were all terrified of retaliation,” Rivers said. “The second we file, we put a big target on our back.”
She had another reason to worry: She was not yet public about her transition and worried that the company would out her. “I was extremely afraid,” she said. “I feared any investigation would reveal communication that used my new name, Rebecca Rivers, rather than my dead name.”
Then in November, Rivers and Berland were abruptly put on administrative leave. Rivers turned up at work one morning and realized her security badge didn’t work. “I called the badging office, and they didn’t know; they reactivated me,” she said. “It was a very rushed turn of events.”
When Rivers and Berland were interviewed soon after, Rivers for the second time, they were asked about material they had accessed, but investigators once again spent much of the interviews on other topics. An investigator asked Rivers a series of questions about her role in the anti-C.B.P. activism, including her social media commentary. “He was just fishing for more information related to C.B.P.,” she said.
According to Berland, investigators asked him about his subscription to the calendars of several senior Google employees who had met with IRI Consultants. They asked if he planned to show up unannounced to any of the meetings, which made no sense to him. “I honestly couldn’t believe they seriously suggested that,” he said. “Until they asked, it hadn’t even occurred to me. Why would I do that?” He said he had subscribed to the calendars, which were open to any employee, because he worried the consultants might be helping the company curtail labor rights.
Within Google, the suspensions generated angst, and even nonactivists began to wonder if the company was overreacting. On Nov. 12, Walker posted an internal note defending the move. He said, apparently alluding to Rivers, that one employee had deliberately “searched for, accessed and shared a number of confidential or need-to-know documents outside the scope of their job” and that another employee’s calendar-tracking had caused “a lot of stress for people who are just trying to go about their work.”
“We have the potential to develop incredible products and services that can help billions of people around the world,” he added, but “we can’t let internal wrangling get in the way of that mission.”
Google fired the four activists on the Monday before Thanksgiving. In an internal memo that soon leaked to the press, the company said the workers were involved in “systematic” searches for material “outside the scope of their jobs.” It also cited an employee’s efforts to track information on the calendars of colleagues “outside of their work group.” And it said that information the employees accessed made its way outside the company — a suggestion that they had abetted leaking at a company where even rank-and-file workers regard leaks as a form of betrayal.
The company later said the workers had accessed need-to-know documents, only some of which were labeled such. All four workers said they were unaware of having inappropriately accessed documents labeled need-to-know, and none were ever accused of directly leaking information. When asked, Google could not point to a specific rule that forbade accessing material on colleagues’ open calendars or setting up calendar notifications. Laurie Burgess, a lawyer representing the workers, said that they had done what Googlers have traditionally been encouraged to do and that the firing was intended to chill employee activism. “They picked on people all over the country — they fired them at the same time,” Burgess said. “It’s sort of giving a lesson to employees everywhere that you could be next.”
Among activists, the reaction was sheer terror. “The folks whose names are on the bottom of the petitions were checking their bank accounts,” said Colby Jordan, a sometime activist based in Seattle who left the company in January. .‘If I don’t have a job, how’s this going to work for me?.” But the firings did appear to mobilize hundreds if not thousands of Googlers who had previously stood on the sidelines. An activist said one city’s weekly lunch, which usually attracted about 100 people, had doubled in size the week after the firings, and an additional meeting on the topic drew 300 people. “It’s a very short-term play,” Jordan said. “When a company is its workers — a fact that’s true in tech in particular — you can’t beat them into submission.”
The more aggressive reaction by the company also began to convince many activists that only a formal union could protect them. Bruce Hahne, a project manager who left Google in February after 14 years, said the crackdown with the Zurich team was a galvanizing moment that created a backlash and drove employees to say: “Let’s learn more about this labor union stuff. What is it that they’re trying to suppress?”
Some Google workers had previously reached out to unions, but the idea was met with little enthusiasm from their colleagues. The success of any worker organization at Google would depend on the number of workers it could mobilize, they said, so why not focus on that? For a while, one popular idea was a so-called solidarity union — a union that didn’t seek certification under federal labor law. Unlike a formally certified union, creating a solidarity union wouldn’t depend on winning a majority of employees in a secret-ballot election, which at Google could theoretically require winning tens of thousands of votes.
But since the firings, Google workers have begun softening on organized labor. “I have felt comfortable within some of the email lists I have posted to, to say here is some labor news or some union news,” Hahne said. “You can say that and not have 20 people jump into the thread telling anti-union stories.”
Some workers have reached out to the Office and Professional Employees International Union, which represents white-collar workers like nurses, podiatrists, clerical staff and university staff, and the Communications Workers of America, which represents telephone-company workers, airline workers and many journalists. “We at C.W.A. are happy to be a resource to folks who are looking for resources to support them,” said Tom Smith, the union’s organizing director. “We want to be known as a place folks can turn.”
Beyond the cultural hurdles of trying to bring traditional unions to an industry in which they have been largely absent, the effort would face significant procedural hurdles as well, not least of which is simply determining who would be eligible to join. Some union proponents, taking note of a group of about 80 Google contract workers from Pittsburgh who had unionized in September, have begun to sketch out plans to circumvent the logistical challenges of unionizing by proceeding office by office, job function by job function.
A young security engineer named Kathryn Spiers epitomized the recent shift in thinking. She had once believed that Google was too large and too scattered geographically to unionize all at once. But over drinks with co-workers to celebrate her 21st birthday, she began to wonder: What if their team of about 100 workers unionized on its own? The conversation sparked interest among her co-workers. “I ended up having to put an F.A.Q. doc together because every conversation had the same few questions, mostly around: Why do well-paid, well-benefited engineers need this? Will this result in more administrative overhead with no real benefit?” she said. “A union will make it easier, especially for security, to stand up when they think what they’re being asked to build is wrong.”
Spiers had not hid her organizing conversations from her managers, frequently chatting with co-workers in their open office space. The ax fell shortly after she created a notification reminding co-workers of their right to participate in protected activities like organizing when they visited the website of IRI Consultants or an internal site about company guidelines. Co-workers could see the notification only if they were using Google-managed devices within the corporate network. “I can’t imagine something more clearly protected,” said Burgess, who is now representing Spiers.
Spiers said she was told only that she violated several policies, but not the specific infractions. (The Google spokeswoman said the notification had been created by misusing internal tools and that the message per se was not an issue.)
In January, hundreds of workers at Amazon began to defy a corporate-communicationspolicy by publishing quotes critical of the company’s business practices. Many in the tech industry saw the protest as part of a wave that had radiated out from Google. “Google’s walkout was huge for us,” said Weston Fribley, an Amazon software engineer involved in climate protests there. “It really expanded our imagination about what was possible.”
Even after the events of the past year, it’s hard to say whether Google’s reputation has been meaningfully tarnished within the broader pool of talented, idealistic engineers it hopes to recruit and retain. After all, it just has to look more ethical and worker-friendly than its rivals, all of which are confronting controversies of their own. As if to hammer home that fact, the day after the Amazon workers’ protest began, a Google recruiter praised the Amazon activists in a LinkedIn post, flagged by Vice. He urged them to consider coming to Google, where they could better live out their values. The post, which soon disappeared, seemed both strangely oblivious to the labor turmoil in Mountain View and inadvertently revealing about the calculations that linger over it. (Google said the recruiter was acting on his own initiative.)
If Google continues to set the curve in the competition to be the most righteous tech employer, then it may feel it can keep cracking down. “There are a small number of firms that could possibly absorb the vast majority of people hired here,” one Google activist said. “It’s like: ‘Go ahead, work for Bezos. Or work for Salesforce. They have a charismatic liberal founder who also works with ICE.”’
On the other hand, the risk of irreparably damaging the company’s worker-swaddling reputation almost certainly constrains Google management, which presumably wants to stop the discontent from leaching further into its work force, to say nothing of the popular consciousness. The shift to a more heavy-handed approach also creates the possibility of a costly overreach. “As recently as six months ago, I thought I’d seen this movie before,” said Hyde, the labor law expert at Rutgers Law School. “Then, for reasons I don’t know but would love to know, the ball seems to have been passed to idiots who seem to be doing everything possible to bring about what their predecessors were able to avoid.”
Perhaps this concern was why, two months after the company announced it was turning the weekly T.G.I.F. ritual into a monthly meeting restricted to a narrow set of questions, Pichai told workers he was reconsidering. At a staff meeting in January, he said that he had heard from many Googlers that T.G.I.F. was important to them and that he would look into finding a way to preserve the original free-for-all concept, according to two people who watched. (Google said it now allows a few off-topic questions at the end of each session.) In mid-February, Google also announced that Naughton, its longtime head of human resources, would leave her position for another role at the company later this year.
The activists are still pushing. In the most recent annual employee survey, many say they vented their frustration over the company’s need-to-know policy, whose enforcement Walker recently referred to at a meeting as “common sense.” There is also the matter of the five workers Google recently fired. In December, the C.W.A. filed charges on their behalf with the National Labor Relations Board, accusing the company of retaliating over their workplace activism. The agency will decide whether to issue complaints, which are akin to indictments, and could eventually reinstate them.
If that happens — and Hyde said that redressing illegal firings is one of the few areas of labor law that workers can still leverage, even with a hostile company and the current administration in Washington — it would deal a blow to the narrative Google has labored to create internally. Rather than viewing their fired colleagues as rogue agents intent on sabotaging the company, many Googlers who currently accept the company’s good faith may begin to question its motives.
And then there is the Customs and Border Protection relationship. It’s unclear when the agency will formally open the bidding on the contract that it indicated in July would be forthcoming. But if the company submits a proposal, the activists will have vulnerabilities to exploit. Not least is Kurian’s remark that the infrastructure would not support the agency’s work on the Mexican border, a claim that even some nonactivists consider dubious.
In the meantime, the surviving Google activists lurch between gallows humor and defiance. With every rumor of a retaliatory firing, dozens of Googlers steel themselves emotionally for what they expect will be the end of the line. But they also believe that there are structural forces propelling their activism forward. “Locally pessimistic, globally optimistic,” was how one put it, descending into engineering speak.
Mar Hicks, a labor historian and author of the book “Programmed Inequality,” said there were in fact reasons to believe that the Google activists could succeed at building a lasting worker movement. For one thing, they said, today’s tech workers, unlike earlier generations, have built coalitions beyond the walls of their own company, making it more difficult for employers to isolate them. They have forged alliances across class and job categories, and they situate themselves within a larger political and cultural reaction to the extreme inequality that the tech industry has come to epitomize.
The uprising also comes at a moment of unprecedented scrutiny of technology in the political realm and in the media, which dovetails with the workers’ own anxieties about how the company monetizes their work. “The potential for top-down regulatory action at the same time that we’re having a large groundswell from the bottom up of labor activism,” Hicks said, “puts us in a somewhat different place than earlier moments.”
Yet even these developments can’t quite explain the workers’ audacious — some might say Googley — sense of confidence. While they recognize their mission to be a moonshot, they may simply be that rare class of worker used to engineering moon landings. “Management can decide whether or not to get on board,” Berland said. “But Google workers are increasingly aware of the power they have. They’re going to continue to exercise that power, and in the end, they’re going to prevail.”