Google bosses have forced employees to delete a confidential memo circulating inside the company that revealed explosive details about a plan to launch a censored search engine in China, The Intercept has learned.
The memo, authored by a Google engineer who was asked to work on the project, disclosed that the search system, codenamed Dragonfly, would require users to log in to perform searches, track their location — and share the resulting history with a Chinese partner who would have “unilateral access” to the data.
The memo was shared earlier this month among a group of Google employees who have been organizing internal protests over the censored search system, which has been designed to remove content that China’s authoritarian Communist Party regime views as sensitive, such as information about democracy, human rights, and peaceful protest.
According to three sources familiar with the incident, Google leadership discovered the memo and were furious that secret details about the China censorship were being passed between employees who were not supposed to have any knowledge about it. Subsequently, Google human resources personnel emailed employees who were believed to have accessed or saved copies of the memo and ordered them to immediately delete it from their computers. Emails demanding deletion of the memo contained “pixel trackers” that notified human resource managers when their messages had been read, recipients determined.
The Dragonfly memo reveals that a prototype of the censored search engine was being developed as an app for both Android and iOS devices, and would force users to sign in so they could use the service. The memo confirms, as The Intercept first reported last week, that users’ searches would be associated with their personal phone number. The memo adds that Chinese users’ movements would also be stored, along with the IP address of their device and links they clicked on. It accuses developers working on the project of creating “spying tools” for the Chinese government to monitor its citizens.
People’s search histories, location information, and other private data would be sent out of China to a database in Taiwan, the memo states. But the data would also be provided to employees of a Chinese company who would be granted “unilateral access” to the system.
To launch the censored search engine, Google set up a “joint venture” partnership with an unnamed Chinese company. The search engine will “blacklist sensitive queries” so that “no results will be shown” at all when people enter certain words or phrases, according to documents seen by The Intercept. Blacklisted search terms on a prototype of the search engine include “human rights,” “student protest,” and “Nobel Prize” in Mandarin, said sources familiar with the project.
According to the memo, aside from being able to access users’ search data, the Chinese partner company could add to the censorship blacklists: It would be able to “selectively edit search result pages … unilaterally, and with few controls seemingly in place.”
That a Chinese company would maintain a copy of users’ search data means that, by extension, the data would be accessible to Chinese authorities, who have broad powers to obtain information that is held or processed on the country’s mainland. A central concern human rights groups have expressed about Dragonfly is that it could place users at risk of Chinese government surveillance — and any person in China searching for blacklisted words or phrases could find themselves interrogated or detained. Chinese authorities are well-known for routinely targeting critics, activists, and journalists.
“It’s alarming to hear that such information will be stored and, potentially, easily shared with the Chinese authorities,” said Patrick Poon, a Hong Kong-based researcher with the human rights group Amnesty International. “It will completely put users’ privacy and safety at risk. Google needs to immediately explain if the app will involve such arrangements. It’s time to give the public full transparency of the project.”
On August 16, two weeks after The Intercept revealed the Dragonfly plan, Google CEO Sundar Pichai told the company’s employees that the China plan was in its “early stages” and “exploratory.” However, employees working on the censored search engine were instructed in late July, days before the project was publicly exposed, that they should prepare to get it into a “launch-ready state” to roll out within weeks, pending approval from officials in Beijing.
“It will completely put users’ privacy and safety at risk.”
The memo raises new questions about Pichai’s claim that the project was not well-developed. Information stored on the company’s internal networks about Dragonfly “paints a very different picture,” it says. “The statement from our high-level leadership that Dragonfly is just an experiment seems wrong.”
The memo identifies at least 215 employees who appear to have been tasked with working full-time on Dragonfly, a number it says is “larger than many Google projects.” It says that source code associated with the project dates back to May 2017, and “many infrastructure parts predate” that. Moreover, screenshots of the app “show a project in a pretty advanced state,” the memo declares.
Most of the details about the project “have been secret from the start,” the memo says, adding that “after the existence of Dragonfly leaked, engineers working on the project were also quick to hide all of their code.”
The author of the memo said in the document that they were opposed to the China censorship. However, they added, “more than the project itself, I hate the culture of secrecy that has been built around it.”
The memo was first posted September 5 on an internal messaging list set up for Google employees to raise ethical concerns. But the memo was soon scrubbed from the list and individuals who had opened or saved the document were contacted by Google’s human resources department to discuss the matter. The employees were instructed not to share the memo.
Google reportedly maintains an aggressive security and investigation team known as “stopleaks,” which is dedicated to preventing unauthorized disclosures. The team is also said to monitor internal discussions.
“More than the project itself, I hate the culture of secrecy that has been built around it.”
Internal security efforts at Google have ramped up this year as employees have raised ethical concerns around a range of new company projects. Following the revelation by Gizmodo and The Intercept that Google had quietly begun work on a contract with the military last year, known as Project Maven, to develop automated image recognition systems for drone warfare, the communications team moved swiftly to monitor employee activity.
The “stopleaks” team, which coordinates with the internal Google communications department, even began monitoring an internal image board used to post messages based on internet memes, according to one former Google employee, for signs of employee sentiment around the Project Maven contract.
Google’s internal security team consists of a number of former military and law enforcement officials. For example, LinkedIn lists as Google’s head of global investigations Joseph Vincent, whose resume includes work as a high-ranking agent at the U.S. Immigration and Customs Enforcement agency’s Homeland Security Investigations unit. The head of security at Google is Chris Rackow, who has described himself as a former member of the Federal Bureau of Investigation’s hostage rescue team and as a former U.S. Navy SEAL.
For some Google employees, the culture of secrecy at the company clashes directly with its public image around fostering transparency, creating an intolerable work environment.
“Leadership misled engineers working on [Dragonfly] about the nature of their work, depriving them of moral agency,” said a Google employee who read the memo.
Google did not respond to a request for comment on this story.