TikTok owner ByteDance is actively suppressing a large number of so-called “sensitive words,” according to an internal company record.
Internal company filings show that TikTok’s parent company, ByteDance, maintains a library of vocabulary lists that dictate what users of their platforms can and can’t see. While this is standard practice across many social media brands, the Chinese company’s glossary goes beyond the standard moderation blocklists, which focus on child safety and things like hate speech.
Posts about the Chinese government, US-China trade, former President Donald Trump, the persecuted Uyghur minority, and even TikTok competitors like YouTube are subject to surveillance and, in many cases, suppression by ByteDance.
As revealed by forbesLists in ByteDance’s tool, which tracks “sensitive words” across its platforms, include titles such as:
- “TikTok Vocabulary for Sensitive People”
- “Trump ordered forbidden words”
- “Putin ordered forbidden words”
- “TikTok Japanese comments suppress words”
- “TikTok audio-sensitive words in the Tibet region”
- “Special Banned Words for Xi and Peng”
- “Sensitive Words in Douyin Videos in Xinjiang”
- “Theming strategies of Uyghur-Hanish couples”
- “Thematic Strategy for China’s Strategic Politics”
- “Local Life – Independence of Taiwan and Hong Kong”
- “YouTube Domestic Surveillance”
- “Negative-Sensitive Vocabulary for Enterprise Products”
- “TikTok Government Affairs Media Topic Vocabulary”
Forbes reports more than 50 listings in the ByteDance tool with the word “TikTok” or “US” in the title. Internal materials show that TikTok employees have accessed the monitoring tool within the last year. Other lists point to TikTok’s Chinese counterpart Douyin and ByteDance products Lark, Toutiao, and Resso.
An internal guide to the tool appears to have been written by China-based employees of ByteDance and its Beijing subsidiary Jiyun Hudong. This guide describes its “core global vocabulary” with “common, high-risk, bottom line sensitive words” and a “global, commonly used thesaurus” containing “fully classified illegal vocabulary”. The document states that the system was developed in 2017 to help “recognize, score and retrieve these words across ByteDance products” as they “pose risks to safety, reputation and revenue.” of the company” and concludes that “the business side will take precedence”. ”
Jamie Favazza, a spokesperson for TikTok, pointed out that the extensive glossary obtained by Forbes could be “significantly out of date or incomplete” and that none of the identified word lists are currently, or have ever been, used on TikTok. When asked why these lists have “TikTok” or “US” in the titles when they’ve never been used on the platform, Favazza said, “I can’t speculate on the list titles” without Forbes releasing the documents.
Hundreds of documents reviewed by Forbes show that internally there is no functional separation in accessing information, user data and tools between TikTok and its Chinese counterpart Douyin. Additionally, dozens of current and former employees of TikTok and ByteDance have revealed that any alleged split between the companies is primarily cosmetic.
In addition to TikTok employees having access to the ByteDance tool over the past year, documents show that ByteDance employees in China are also among those who managed the TikTok tool. Still, according to Favazza, TikTok could not verify if any of its employees accessed the tool without knowing their names.
Records show that these moderation tools and other internal programs collect data about the “match rate” of sensitive words – including information about US users who post them. A document mentions that TikTok and other ByteDance products recently received an upgrade that integrated the moderation system with “a new text recognition service” that makes it easier to track and analyze “sensitive word hit records” in real time. At least one of the engineers leading this project was based in Beijing.
Forbes’ findings also revealed extensive overlap between Chinese state media and ByteDance/TikTok staff. Chinese state media have already used TikTok to try to sway American users’ opinions of US politics, despite TikTok CEO Shou Zi Chew testifying before Congress last month that there was “insufficient” evidence to “prove or disprove, that TikTok uses political censorship”.
“The difference is that ByteDance is subordinate to the government and the Department of State Security,” says William Evanina, a former US government counterintelligence chief. “Yes, Facebook and Google are all doing the same to protect their global assets on their platforms, but they are under no obligation to the CIA or NSA. And the NSA and CIA don’t share that with Facebook (Glossary).”
From a cybersecurity perspective, experts cite the danger of ByteDance’s surveillance system: the ability to track how many times certain words appear, who said or interacted with them, where those people are located, and who else is following them.
In China, laws require companies to provide information to the government upon request. As such, ByteDance could be required to share sensitive information with party leaders — whether that activity originated in China, the US, or another country.