In a 2016 address to propaganda cadres and state-run media personnel, Chinese President Xi Jinping expressed dreams of instilling a new international media order “wherever the readers are, wherever the viewers are; that is where propaganda reports must extend their tentacles.” As Xinhua News, China’s largest state-run news agency, equips itself with “Media Brain,” an artificial intelligence (AI) newsroom to assist all stages of reporting, these “tentacles” of propaganda may extend faster.
Bringing AI to newsrooms can improve accuracy, enhance data analysis, and increase efficiency. According to a video released by Xinhua in January, the AI newsroom will do everything “from finding leads to news gathering, editing, distribution, and, finally, feedback analysis.” Last week, Xinhua announced an update to Media Brain called “MAGIC,” which will use machine generated content (MGC) for “fast-speed news production” and can automatically generate a news video in as fast as 10 seconds.
Cai Mingzhao, president of Xinhua and member of the State Council Information Office, said the company wants to use AI to “realize individualized and customized information delivery.” As technology improves, “customized information” will take many forms, from personalized digital news feeds to headlines and articles tailored to individual readers—and, likely, to propaganda. Xinhua did not respond to request for comment.
While Xinhua is discussing its AI ambitions in the context of improving user experience, the state-run agency first and foremost serves the Chinese Communist Party (CCP). Adding AI into the mix could have insidious consequences for truth-seeking individuals in China and in the numerous countries around the world where Xinhua shares its content.
“It’s mostly about improving propaganda,” says Sarah Cook, East Asia analyst at Freedom House, a civil liberties advocacy group, of the agency’s AI initiative. “If they’re able to reach more people with more convincing propaganda then that affects how people view the world, how people view China.”
SOMETIMES CALLED “the world’s largest propaganda agency,” Xinhua News receives information directly from the Chinese Communist Party (CCP), and generates newswires that fuel journalism across mainland China. “Its content is spread throughout pretty much any other form of reporting inside China,” says Cook. “So you might pick up a newspaper or go online, and there might be a combination of original reporting from more commercialized outlets in China, with Xinhua sprinkled in.”
Xinhua also has a growing international presence, with more than 170 bureaus abroad and the largest number of foreign correspondents from any news agency around the world. “In the past, Xinhua looked up to Reuters and AP. Xinhua wanted to catch up,” says Xin Xin, a former Xinhua journalist and reader in International Communications at the University of Westminster. “But suddenly, Xinhua realized it was still expanding while other news agencies were downsizing.” Of course, as a state-run agency, Xinhua does not rely on advertising revenues.
An AP writer is going to write one article…but what if you could write a hundred? Now, you’re not just a major player in the already-small number of players controlling a pretty big influence in a global conversation, but you’re the best player.
Sean Gourley, founder and CEO of Primer, a San Francisco-based machine intelligence company, points out that Xinhua could join the large newswires in the ranks quickly—and gain a lot of influence. “You’ve got AP, Reuters, somewhat CNN wire, you’ve got a few others, but there aren’t many, and these [newswires] just reverberate massively,” he says. “News agencies have a huge, huge role setting the tone of the conversation, much, much more than we give them credit for.”
Through information exchange agreements with news outlets in countries as far-ranging as Argentina and Kenya, Xinhua is already spreading its content across continents. And with AI on Xinhua’s horizon, the CCP’s narrative will certainly have greater reach. All newswires that use AI will. “Once you start [writing wires] in an automated fashion, all of a sudden, you can scale up very, very quickly,” says Gourley. “An AP writer is going to write one article…but what if you could write a hundred? Now, you’re not just a major player in the already-small number of players controlling a pretty big influence in a global conversation, but you’re the best player.”
It looks like Xinhua might soon have a leg-up on the competition. In last week’s announcement, the agency predicted the updated Media Brain could generate over 10,000 news videos during this year’s World Cup.
Should AI amplify Xinhua’s voice outside of China, the agency could have a dramatic global impact on how news spreads. The state-run outlet is infamous for its banned words, blind-spots to human rights issues, accusations of espionage, and, most recently, for making The New York Times’s list of Twitter accounts who purchased fake followers. In 2016, after Xinhua published what was believed to be a forced confession from a reportedly kidnapped book publisher, Gui Minhai, Reporters Without Borders called for EU sanctions against the agency, denouncing it as a “mass propaganda weapon” that “cease[s] de facto to be news media.”
Beyond blocking words or excluding content from news, the government carefully controls the media through requiring use of Xinhua’s content. “The first priority of the CCP is infusing positive narratives about China, telling the ‘good China story,’” explains Cook. “It is about presenting communist government and authoritarian regime in a positive light, portraying them as benign; downplaying human-rights problems; playing up economic successes; and then, at the same time, suppressing critical voices and, when needed, using these avenues to vilify critics.”
Cook calls it “positive censorship.” If negative events or politically sensitive issues like rising tensions in the South China Sea or Taiwanese independence are covered at all, the government will often require use of Xinhua’s copy. “You’ll have the same report, basically, across all the different newspapers or websites, and it’s exactly the same wording,” she says. If the infamous Sinclair episode that sparked outrage across the US had happened in China, it would have been just another Monday.
TO SUCCESSFULLY TAKE the world stage, some say Xinhua will need a makeover. Daya Thussu, an advisor to the China Media Centre, believes China’s growing global influence will enable Xinhua to become a widely cited source for international news. But first, he says, “the news agency will have to go beyond its rather bland and bureaucratic output and make its content interesting and engaging.”
Xinhua has been building hype around the new AI technology as a key player in the “era of augmented journalism.” In one video, European Xinhua correspondent Helen Bentley emphasizes how Media Brain will better respond to people. “It’s no longer just about top news anymore,” she says. “The next step is to identify the concerns of the people, find out what they think.”
After a survey found that 76 percent of Chinese respondents were worried about how AI would impact their privacy, Xinhua released another video stressing the continued importance of reporters alongside the new technology. In the video, strolling through the brightly lit, open-plan office, Bentley explains how the Media Brain collected data to assist reporters covering the National People’s Congress in March.
If the infamous Sinclair episode that sparked outrage across the US had happened in China, it would have been just another Monday.
At a first glance, the aspects of AI Bentley focuses on seem like no more than a supercharged search engine, providing a data dashboard for journalists to maximize their productivity and expand the breadth of their coverage. From Xinhua’s press release, we know plans for the AI newsroom go beyond this data-gathering tool to “customized and individualized information delivery.”
AI could eventually be used to generate custom headlines. Say Xinhua’s Media Brain reports on a woman being robbed in the street. If the algorithm knows the reader is also a woman, it might generate a gender-focused headline. If the reader is a banker, the headline could focus on the amount of money taken. A huge fan of action movies? Perhaps a more sensational headline related to the violence would pop up. The possibilities are endless. It would be like Facebook ad-targeting, international media-style.
In China, this feature could allow the government to emphasize certain elements of a story to specific members of society. If, in the aforementioned example, the robber were of the Uighur minority group, the headline generated for Han majority readers might focus on ethnicity and act as a way of justifying the harsh surveillance practices in Xinjiang province, where Uighurs live. Simultaneously, ethnic minority readers, who may be more likely to experience surveillance in their daily lives, might see a headline that focuses on unpatrolled streets leading to crime.
When stories don’t fit the CCP’s preferred narrative, Xinhua has historically worked to sow seeds of doubt about the validity of those stories. In 1989, the Chinese government used tanks to end protests in Tiananmen square, killed hundreds if not thousands of people, and refused to release a death count. Last summer, when the US called attention to the 28th anniversary of the massacre, Xinhua responded by accusing the US of ignoring objective facts and interfering with internal affairs.
Gourley suggests the Media Brain could automatize these interjections and rebuttals from the government via Xinhua, allowing doubt to spread more quickly, more precisely, and to even more conversations through Xinhua’s footholds in other countries.
In the foreseeable future—from a year to 18 months, Gourley predicts—AI will be able to frame stories in real time. “The idea that there’s one article for everyone is going to quickly change to the one article for me,” says Gourley. “That is something I’m not sure we’re ready for.”
AS XINHUA INTEGRATES AI into the newsroom, the agency is expanding its“media co-operations,” where Xinhua provides its content and, more recently, its technology to foreign outlets for free. Often, these outlets are in countries with a scarcity of journalistic resources and Xinhua’s reporting can fill gaps in coverage. Chinese media scholars call it “borrowing the boat to reach the sea.” Xinhua has been expanding these co-operations, and with AI content could spread through these channels much faster.
“That type of modeling in and of itself might not be problematic,” says Cook in regards to Xinhua sharing its technology with reporters, “but it affects how people see Xinhua and therefore use their content.” If people learn about technology from Xinhua, there may be a sense of obligation to use more of its content, Cook suggests.
The idea that there’s one article for everyone is going to quickly change to the one article for me. That is something I’m not sure we’re ready for.
Just having AI in the newsroom could make Xinhua competitive in the global media market.
“As someone who has worked in media, there is a strong emphasis on being first when it comes to experimenting,” says Jarrod Dicker, formerly head of a digital research lab at The Washington Post, about the use of AI in newsrooms worldwide. “It’s about seeing if this works before anyone else.” It’s not just Xinhua who wants to be first; other news organizations are also integrating AI into their operations—including AP and Reuters.
With the addition of the Media Brain to Xinhua’s newsroom, the AI arms race has taken on greater significance for journalism as a race for information. Bolstered by China’s heavy investment in AI and edged on by the CCP’s control of the media, Xinhua could gain headway quickly.
“This is a very, very powerful organization powered by increasingly sophisticated technology,” says Gourley. “I think it is going to imprint the Chinese perspective on the world.”