For their MTF project, the team which was formerly known as Tracking.Exposed (now known as AI Forensics) proposed a set of tools which would allow for the collection and analysis of TikTok user data in real-time. The TikTok Observatory was among the first tools of its kind and was modeled after similar tools that AI Forensics had built for other platforms like Facebook. Mozilla was the first funder to support this project, placing a “risky bet” to support the development of a utility that could be used to audit and hold accountable a company that was becoming increasingly dominant in the global social media sphere. Unlike other social media platforms, TikTok does not provide a public API, so researchers were forced to build their own tools to be able to analyze the network’s behavior.
“The TikTok Observatory developed the infrastructure needed to investigate recommended and demoted content on the ‘For You’ feed — that was the original idea,” explained Marc Faddoul, co-director of AI Forensics. TikTok’s shadow-banning (e.g. censorship that is invisible to the user) of content the platform deemed undesirable was already known from a 2020 Intercept report; the AI Forensics team was curious to see if shadow-banning was being used to suppress other topics. When Russia invaded Ukraine in February of 2022, the AI Forensics team was quietly watching TikTok’s behavior through their collection of data in real-time.
What the team found was shocking. In a bombshell report published in March 2022—less than a month after the start of the war—AI Forensics reported that 95% of content was blocked for TikTok users in Russia. This was especially noteworthy because prior to the war, TikTok was one of the last sources of uncensored content still available inside of Russia. The team would go on to publish two more landmark reports in 2022, one showing that pro-war content was now dominant on the platform in Russia and another showing that TikTok appeared to be promoting pro-war content to its users. AI Forensics’ work received coverage in a variety of news outlets (including the Washington Post) and was cited as evidence when TikTok’s CEO was called to testify in front of the U.S. Congress. Ultimately, TikTok decided to change its behavior as a result of AI Forensics’ investigations—a massive impact resulting from the hard work of a small and dedicated team and $50,000 USD in support from the Mozilla Technology Fund.
“MTF 2022 allowed us to build auditing free-software to scrutinize TikTok's content moderation in Russia…Our reports forced the platform to fix its policy!” the team wrote. “A rewarding outcome of our second report was that it served as the foundation for a group of U.S. senators to compose a formal letter, summoning TikTok's CEO to explain and fix the platform's content moderation policy in Russia. This was a direct impact of our work.” What’s more, the MTF award set up the AI Forensics team to be able to scale their organization and impact in the long-run. “The project grew the recognition of our organization as experts in the field of algorithmic auditing (for instance, thanks to the media coverage),” the AI Forensics team wrote. “The compelling results and impact we were able to achieve have also proven very helpful in our fundraising efforts.”