TikTok opened a transparency center as it faces renewed threats of government bans

This Tuesday, after a recent charm offensive in Washington, DC, TikTok hosted journalists at its Los Angeles headquarters to open a new center created to attract policymakers, regulators, and American civil society leaders.

“How much of a national security threat is it to join the wifi network here?” NPR tech reporter Bobby Allyn joked as he waited with me and other attendees for the executive presentation to begin. The TikTok staff seemed unsure of what to say until Allyn was sure they were just joking.

The exchange revealed the tension underlying the friendly press invitation: TikTok, the increasingly influential social media app used by more than 130 million Americans, faces intense political scrutiny in the US over its parent company’s ties to China. Less than three years after President Donald Trump tried to ban it, the company’s talks with US regulators have stalled and it faces new calls for a national ban. Currently, 17 US states have banned these apps from government-issued devices.

TikTok’s new Los Angeles Center for Transparency and Accountability offers a behind-the-scenes look at TikTok’s algorithms and content moderation practices, which have drawn controversy amid concerns that the wildly popular app could be used to promote pro-China government messaging or misinformation.

The information that TikTok offers about its algorithms and content moderation is not very clear, but what is telling is the details it shared about its plan to split part of its US operations from China, while still being owned by the Chinese company. The event also provided a rare opportunity for journalists to question a broad cross-section of TikTok staff about its content policies and algorithms.

In an opening speech to reporters, TikTok COO Vanessa Pappas acknowledged the general skepticism about the power of social media platforms in part of our digital lives – without mentioning the specific political problems with TikTok.

“We absolutely understand the criticism,” Pappas said of Big Tech’s role in controlling “how algorithms work, how moderation policies work, and the flow of system data.”

However, Pappas said, TikTok is addressing these concerns by offering what it calls an “unprecedented level of transparency,” with initiatives like the new center and plans to implement other initiatives, such as starting to open TikTok’s API to researchers.

The elephant in the room

There is one big reason we are all in the TikTok office: China. But Pappas and other company leaders never said “China” in their recorded comments.

TikTok is owned by a Chinese company, ByteDance, which operates its own TikTok app, called Douyin, in China.

Critics have long argued that the Chinese-owned company is subject to China’s national security laws, meaning ByteDance employees could be forced to monitor Americans or manipulate TikTok’s recommendation algorithm for Chinese government services. Although there is no evidence that the Chinese government has directly demanded American user data from TikTok or its parent company, an investigative report by BuzzFeed News claims that by June 2022, Chinese TikTok employees may have access to US user data.

At Tuesday’s event, TikTok shared more about its plans to reassure the public that it will not be influenced by the Chinese government. The “Texas Project” is a major partnership with Texas-based tech giant Oracle to move all US data previously stored on TikTok’s foreign servers to the US. The project also required inviting a team of outsiders, including from Oracle, to audit the algorithm.

Another part of the project will create a new subsidiary called TikTok US Data Security (USDS) that will oversee the app’s content moderation policy, train TikTok’s recommendation engine with US user data, and authorize editorial decisions. Under TikTok’s plan, USDS employees will report to an independent board of directors with strong national security and cybersecurity credentials.

This all comes about a month after TikTok was found to be spying on Forbes reporter Emily Baker White, who covered leaked details about the project. TikTok acknowledged that several employees wrongly accessed White’s personal user data, as well as several other journalists, in an effort to identify and track private sources. The company exonerated the employees involved in the surveillance and said they “abused their authority” to obtain user data, but the incident only raised suspicions about the company.

That suspicion could be a factor in why TikTok’s negotiations with the US Committee on Foreign Investment in the US, or CFIUS, are moving forward. CFIUS is an interagency government committee that reviews whether business transactions pose a threat to US national security. CFIUS has reviewed ByteDance’s 2017 merger of TikTok and Musical.ly, giving it the power to scrap the deal and force TikTok to sell to a US company. Both TikTok and CFIUS are reportedly close to an agreement to avoid that scenario, but negotiations have stalled.

It is widely acknowledged that political escalation between China and the US played a role in the delay. This is not a good time for any political establishment or elected officials — including President Biden, who must sign the agreement — to support anything seen as pro-China.

“TikTok has realized that this is a political issue. It’s less about convincing national security authorities and more about convincing politicians,” said Anupam Chander, professor of law and technology at Georgetown University.

Chander was part of a small group of academics, lobbyists, and data privacy experts that TikTok presented about Project Texas in Washington, DC, a few weeks ago. The challenge, Chander said, is that “today, in certain political circles, the relationship with China is poison.”

This may explain why TikTok executives did not mention China on Tuesday.

Going under the hood

TikTok’s new Center for Transparency and Accountability offered reporters details on its elusive recommendation algorithm and some real-world examples of how the app moderates content, but nothing is telling.

One tutorial in the middle is about TikTok’s recommendation algorithm, which it calls a “code simulator.” It explains how the first time you open the app, you will be shown eight videos from trending topics that TikTok thinks you might be interested in. Then, the app refines its understanding of your interests based on what videos you’ve liked, watched, and shared, what accounts you follow, and what people in the same demographic like. The tutorial shows the code snippets used to program the machine learning model that recommends the content.

The second educational exercise – and more fun – is a simulation of moderated controversial content on TikTok. One video showed a man making jittery movements with his arms with a caption saying he had just received a dose of the vaccine – set to a laugh track. Next to the video, a screen detailing TikTok’s misinformation policy. (The video is not in violation because it is considered humor and not false health information.)

The training provided a better understanding of the tough calls TikTok’s more than 10,000 people around the world who work on trust and security must make every day. But I want to know more about the process of creating TikTok’s guidelines and designing the algorithm: Who decides what content will be seen by more people on TikTok, and how does the app decide when to increase or decrease certain content?

TikTok staff told us that the app only promotes .002 percent of videos on the platform, and that the decision is made by the content programming team, which identifies videos that have the potential to trend. One example he gave was how the company manually gave the Rolling Stones a boost when the band first joined TikTok.

TikTok says it gives access to some outside experts to more detailed details: all the source code, as well as specifics about the exceptions made manually for the promotion of certain trending content, in a separate secret room in Maryland (you have to sign an NDA to enter). The company also said Oracle employees have reviewed TikTok’s code at a separate transparency center in Maryland.

While TikTok’s transparency center sheds more light on how the company and its app work, many still don’t know much about how content, data, and moderation decisions work within the company.

On the other hand, TikTok is taking some new approaches to try to explain its data and algorithmic practices. Under TikTok’s USDS plan, a group of Oracle employees and security experts are supposed to oversee the company’s proprietary algorithms that dictate what millions of people see every day when they log into the app. We have no external level of responsibility for Facebook or YouTube. Companies like Meta and Google also track large amounts of our personal information online but don’t attract the same kind of national security concerns as TikTok because it’s an American company. Although TikTok is currently sharing information due to political necessity, it is very positive for society that they are sharing any information.

It remains to be seen whether TikTok will be able to change minds on Capitol Hill. While this latest initiative is a first step, it will take more — and the validation of outside partners and experts — to persuade TikTok’s staunchest skeptics.