Hayward: Google Builds Censored Search Engine for China But Refuses A.I. Work for U.S. Military

Li Xin/AFP/Getty Images
Li Xin/AFP/Getty Images

Google plans to launch a censorship-friendly search engine for the Chinese market, tailored to the demands of China’s authoritarian communist government.

That is quite a contrast with Google’s reluctance to work for the U.S. military. Just a few months ago, an employee revolt at the tech giant nearly killed a vital artificial-intelligence contract with the Pentagon.

The Intercept on August 1 cited internal Google documents to tell the tale of “Dragonfly,” a project launched last year and “accelerated” after a productive meeting between Google CEO Sundar Pichai and a Chinese official last December.

Dragonfly sounds like a totalitarian wet dream, an Orwellian nightmare built from the ground up to satisfy China’s hunger for speech control:

The Chinese government blocks information on the internet about political opponents, free speech, sex, news, and academic studies. It bans websites about the 1989 Tiananmen Square massacre, for instance, and references to “anticommunism” and “dissidents.” Mentions of books that negatively portray authoritarian governments, like George Orwell’s 1984 and Animal Farm, have been prohibited on Weibo, a Chinese social media website. The country also censors popular Western social media sites like Instagram, Facebook, and Twitter, as well as American news organizations such as the New York Times and the Wall Street Journal.

Documents seen by The Intercept, marked “Google confidential,” say that Google’s Chinese search app will automatically identify and filter websites blocked by the Great Firewall. When a person carries out a search, banned websites will be removed from the first page of results, and a disclaimer will be displayed stating that “some results may have been removed due to statutory requirements.” Examples cited in the documents of websites that will be subject to the censorship include those of British news broadcaster BBC and the online encyclopedia Wikipedia.

The search app will also “blacklist sensitive queries” so that “no results will be shown” at all when people enter certain words or phrases, the documents state. The censorship will apply across the platform: Google’s image search, automatic spell check and suggested search features will incorporate the blacklists, meaning that they will not recommend people information or photographs the government has banned.

Within Google, knowledge about Dragonfly has been restricted to just a few hundred members of the internet giant’s 88,000-strong workforce, said a source with knowledge of the project. The source spoke to The Intercept on condition of anonymity, as they were not authorized to contact the media. The source said that they had moral and ethical concerns about Google’s role in the censorship, which is being planned by a handful of top executives and managers at the company with no public scrutiny.

Contrast this with the enormous controversy over Project Maven, a Pentagon program to develop artificial intelligence that can assist with drone strike targeting. Over 3,100 Google employees, including some of its senior engineers, signed a letter in April demanding the end of the project because “Google should not be in the business of war.” A number of Google employees quit their jobs to protest the contract.

The Google letter dismissed promises that Maven technology would not be employed to directly pilot the drones or launch their weapons: “While this eliminates a narrow set of direct applications, the technology is being built for the military, and once it’s delivered it could easily be used to assist in these tasks.”

The letter asserted a moral position that would become a sick joke if Google helps the tyrants of Beijing oppress their people by distorting and censoring what they see online:

This plan will irreparably damage Google’s brand and its ability to compete for talent. Amid growing fears of biased and weaponized A.I., Google is already struggling to keep the public’s trust. By entering into this contract, Google will join the ranks of companies like Palantir, Raytheon, and General Dynamics. The argument that other firms, like Microsoft and Amazon, are also participating doesn’t make this any less risky for Google. Google’s unique history, its motto Don’t Be Evil, and its direct reach into the lives of billions of users set it apart.

We cannot outsource the moral responsibility of our technologies to third parties. Google’s stated values make this clear: Every one of our users is trusting us. Never jeopardize that. Ever. This contract puts Google’s reputation at risk and stands in direct opposition to our core values. Building this technology to assist the U.S. government in military surveillance – and potentially lethal outcomes – is not acceptable.

Evidently, playing footsie with the oppressive and murderous regime in Beijing is acceptable, if that is what it takes to give Google a crack at the huge Chinese marketplace.

Google managers announced in June that the company would not renew its contract with the U.S. military. The company decided not to abandon all work for the military but published a set of guidelines for A.I. that would prevent working on anything like Maven.

The guidelines laid out by Pichar in a June 7 post expressly forbid working on “technologies that cause or are likely to cause overall harm,” weapons, surveillance tech that would violate “internationally accepted norms,” and technology that “contravenes widely accepted principles of international law and human rights.”

Those last two guidelines can apparently be disregarded if the Chinese government wants sinister surveillance tech and speech controls that violate every single principle of human rights, and the reward for compliance is access to Chinese markets.

The ethical compromises behind the Dragonfly project have not gone unnoticed by Google employees. 1,400 of them recently signed a letter protesting how the China search engine project has been handled, complaining in particular about the way rank-and-file engineers have been kept in the dark:

Our industry has entered a new era of ethical responsibility: the choices we make matter on a global scale. Yet most of us only learned about Project Dragonfly through news reports in early August. 

Dragonfly is reported to be an effort to provide search and personalized mobile news to China, in compliance with Chinese government censorship and surveillance requirements. Eight years ago, as Google pulled censored websearch out of China, Sergey Brin explained the decision, saying “in some aspects of [government] policy, particularly with respect to censorship, with respect to surveillance of dissidents, I see some earmarks of totalitarianism.”

Dragonfly and Google’s return to China raise urgent moral and ethical issues, the substance of which we are discussing elsewhere.

China sees stories like this as tributes to its “sharp power” strategy of using market access as a club against Western corporations. Beijing is increasingly bold in using its tightly-controlled markets as ideological weapons, forcing American and European companies to abandon their commitment to free expression and accept Chinese speech codes.

This is a form of economic warfare China loves because it believes Western societies have little defense against it, and cannot retaliate in kind. There is little the U.S. government could do that would compensate a corporation for suffering a politically-motivated ban from Chinese markets. There is zero chance any Western government would use command economics to force Chinese business interests to obey their ideals.

Does anyone want to gamble that China cannot use sharp power to make companies like Google quietly abandon their principles against weaponized A.I. and assist Chinese military projects?

Even if it never comes to that, the race for A.I. is red hot, and the United States is not in the lead. Western corporations striking “principled” poses against doing A.I. work for the Pentagon are only helping the most aggressive and amoral power in the world build up its lead.

Chinese engineers will appropriate and weaponize Western tech as needed, and they are designing plenty of cutting-edge systems on their own. Project Maven used A.I. to help the U.S. military track vehicles and equipment, not individual people; the goal was to help monitor and target terrorists more accurately, which would reduce civilian casualties. China is already using A.I. systems to monitor the populations of entire cities.

The next generation of artificial intelligence will be the cyber equivalent of the atomic bomb, a weapon with enormous first-strike potential, a revolutionary impact on battlefield tactics, and stunning implications for economic development. The good guys must develop that technology first. Refusing to help America win the race will objectively help China win, and then all that lofty talk of “doing no harm” and “helping information to be free” will become just another cluster of meaningless words easily censored by the authoritarian masters of the 21st Century.


Please let us know if you're having issues with commenting.