Senators Demand Answers After TikTok Ran Experiment Disabling Safety Feature for 15 Million U.S. Users
U.S. Senators Marsha Blackburn (R-TN) and Richard Blumenthal (D-CT) sent a letter to TikTok CEO Shou Chew and Adam Presser, CEO of TikTok's U.S. business, on August 20, 2026, demanding answers about an experiment that intentionally turned off an algorithmic safeguard for roughly 10% of U.S. users — a control group of approximately 15 million people. The senators called TikTok's decision to run the experiment "depraved" (TechCrunch).
The safeguard in question was designed to prevent users from being flooded by harmful content. TikTok ran the experiment to find out whether the safeguard made the app less engaging, according to a confidential TikTok report that stated an algorithm adjustment was withheld from many users "by design" as part of an effort to measure engagement (Bloomberg). The Bloomberg report was published on August 4, 2026.
The senators' letter reveals that 16-year-old Chase Nasca was in the control group. His TikTok algorithm fed him thousands of videos about sadness, suicide, and loneliness up until he died by suicide (TechCrunch).
Blackburn and Blumenthal posed 13 questions to TikTok, including the names of every employee who knew about the experiment and why TikTok allowed minors to be part of the control group. They also requested a list of every algorithmic experiment conducted in the U.S. where TikTok disabled or delayed a safety feature. TikTok has until September 1 to respond. TikTok did not immediately respond to TechCrunch's request for comment (TechCrunch).
The letter builds on years of congressional scrutiny of TikTok's child-safety practices. In October 2024, two U.S. senators who had pushed online child safety legislation in Congress demanded that TikTok executives produce documents in response to their child-safety demands (NPR). A Senate hearing, numbered 117-843 and titled "Protecting Kids Online: Snapchat, TikTok, and YouTube," was published by the U.S. Government Publishing Office (GPO). The senators' current letter also cites a Wall Street Journal article from December 17, 2021, titled "The Corpse Bride Diet: How TikTok Inundates Teens With Eating-Disorder Videos" (Senate HELP Committee).
Senator Blackburn's engagement with TikTok extends further back. A letter on her official Senate website dated November 26, 2019, noted TikTok's stated commitment to safeguarding the data of its users, especially its younger users (Blackburn Senate). A subsequent letter from June 30, 2022 stated that TikTok's solution under "Project Texas" focused on evaluating and revising TikTok's internal policies (Blackburn Senate).
The broader national-security architecture around TikTok's U.S. operations has also faced recent congressional pressure. Democratic Senator Ed Markey proposed new national security safeguards on July 31, 2025, that would ensure TikTok could remain online (Reuters). On May 29, 2026, Markey pressed TikTok and Oracle on national security concerns with the new TikTok U.S. entity, stating that the first safeguard of the TikTok deal, source code review, raises three serious issues (Markey Senate).
The structure of the experiment TikTok ran is straightforward in A/B-testing terms. A/B testing is a standard product-development technique: you withhold a feature from one group of users, measure how their behavior differs from another group that still has the feature, and use the comparison to decide what to ship. In this case, TikTok withheld a safety feature from a treatment group and measured engagement differences against a control group. The results would tell the company whether the feature's engagement cost justified its protective function.
What makes this instance different from routine product experimentation is the nature of the safeguard being tested. The feature existed specifically to limit exposure to harmful content. Disabling it to measure whether it reduced engagement effectively quantified the trade-off between user safety and session length, and it did so without apparent regard for whether users in the control group, including minors, were exposed to elevated risk as a result.
The Nasca case gives that abstraction a human face. The senators' demand for the names of every employee aware of the experiment signals an escalation beyond policy inquiry into individual accountability, which is unusual for congressional oversight of algorithmic systems.
The broader context here is regulatory. The national-security framework under which TikTok operates in the U.S. — source code review by Oracle, the Project Texas restructuring, and Markey's proposed safeguards — was designed to address data flow and foreign-influence concerns. It was not designed to govern algorithmic experimentation on minors. The Blackburn-Blumenthal letter effectively tests whether the existing oversight apparatus can reach into TikTok's recommendation layer, or whether that gap requires new legislative tools. The September 1 deadline will indicate whether TikTok treats this as a routine inquiry or a contention point worth resisting.
In my view, the core issue is not whether TikTok is technically within its rights to run product experiments. Companies do this constantly across the industry. The issue is that the feature being tested existed to prevent harm, and disabling it for 15 million people — some of them teenagers — to measure the engagement cost of safety is a category of experimentation that most people would assume requires some form of independent review. The fact that it apparently did not is the gap Congress is now probing.
