NextFin News - Two US senators have demanded answers from TikTok over a safety experiment that kept a suicide-prevention feature away from about 15 million users, calling the test "depraved" and sharpening the most costly question now facing the social-media industry: whether platforms priced the risk of protecting children before they would agree to pay it.
Sens. Marsha Blackburn (R-Tenn.) and Richard Blumenthal (D-Conn.), the bipartisan authors of the Kids Online Safety Act, sent the letter to TikTok executives on Wednesday, two weeks after the Senate Commerce Committee unanimously advanced their child-safety bill. The demand follows a report that a confidential internal document showed TikTok withheld an algorithmic safeguard from 10% of US users in 2021 to test whether a safer feed would reduce engagement. The control group included Chase Nasca, a 16-year-old from New York whose account was fed thousands of videos about suicide, sadness and hopelessness before he died by suicide in February 2022.
The episode lands as Congress pushes two competing child-safety regimes, a California jury has begun finding platforms negligent for addictive design, and TikTok has settled a string of teen-suicide cases rather than let any reach a jury. For investors and operators across the attention economy, the question is no longer whether a platform knew harm was possible. It is whether the company measured the revenue cost of preventing it.
The Experiment: Safety as the Variable Under Test
The mechanics matter, because they turn a content-moderation failure into a product-design decision. In 2021, TikTok changed its algorithm to stop users from being flooded with harmful content. But the company did not roll the safer version out to everyone. According to the 14-page internal document, written in March 2023 and held under court seal, TikTok created a control group of 10% of US users — roughly 15 million people at the time — who kept the old version of the app. The stated purpose was to see whether the change might reduce the app's "stickiness," an engagement metric that Wall Street watches closely.
The document says the control group included Nasca's account, and it states plainly why the teenager never received the protection:
"TikTok's filter bubble prevention strategies did not take effect on this user by design."His feed then served him a stream of suicide, self-harm and depressive content. The internal review examined the 7,563 videos Nasca saw in his final two weeks and found that 73% carried themes such as sadness or mental-health struggles without breaking platform rules; nearly 10% did violate policies against content that promotes or normalizes suicide and self-harm. Nasca continued receiving suicide-related videos during the final hours of his life.
The test ran for up to a year. In the document's own recommendations, TikTok proposed shrinking the control group to less than 1% and limiting participation to seven days instead of six months to a year. Even at 1%, the company noted, at least 1.8 million US users would remain without the safeguard.
TikTok told the investigating reporter that the 2022 test was routine and that it has since refined its testing.
"Our hearts break for any family that experienced a tragic loss,"a company spokesperson said. The company said it is "deeply committed to the safety and well-being of users," especially teens, and invests heavily in detection and enforcement.
What TikTok Told Congress — and What It Did Not Say
The political force of the letter comes less from the experiment itself than from the paper trail around it. In written responses to lawmakers in May 2023, TikTok acknowledged using A/B tests to evaluate strategies intended to break up filter bubbles involving suicide and self-harm content. But when asked to list every case in which algorithmic testing had harmed users, the company did not identify any. It wrote that
"the safety of the TikTok community is of the utmost importance"and described how its Community Guidelines applied during testing. It did not mention Nasca or the experiment identified in its own internal review.
That gap is what the senators are now pressing. Blackburn and Blumenthal have built their legislative pitch on a repeated claim: that TikTok's public assurances about teen safety have not matched its internal findings. In an October 2024 letter, the pair wrote that
"TikTok has knowingly designed their products in a manner that can cause substantial harm to kids — including fostering destructive addiction and amplifying child sexual exploitation. Rather than address these risks, TikTok instead seemingly misled the public about the safety of its platform."
TikTok CEO Shou Chew, for his part, told Congress that
"safety and wellness — in particular for teens — is a core priority for TikTok,"and that the company had "a commitment to transparency." Nasca's parents watched from the gallery as lawmakers played examples of videos that had appeared on their son's account.
A former TikTok trust-and-safety team leader who was at the company when the test ran told the investigating reporter he had not known about it. He said A/B tests are common in the industry, but that this one was flawed: the control group was too large, and it should have excluded minors and content such as suicide videos. TikTok did not address those points directly.
The Liability Channel: Why "By Design" Is the Phrase That Matters
Those two words carry weight in court. Thousands of lawsuits argue that social platforms are dangerous "by design," a claim the companies, including TikTok, deny. In March, a California jury found Google's YouTube and Meta's Instagram negligent for targeting children with addictive features, awarding one plaintiff $6 million. TikTok has never let a teen-addiction case reach a jury; it settled three more on the eve of that trial.
The mechanism here is different from the usual moderation complaint. A platform that fails to catch harmful content can argue scale: billions of uploads, imperfect classifiers, an impossible task. A platform that deliberately withholds a working safeguard from a named control group to measure the engagement cost cannot make that argument. The experiment converts an abstract duty of care into a spreadsheet: the company knew the safer version existed, it knew which users would not receive it, and it ran the test long enough to observe the outcome.
This is the transmission channel that turns a safety story into a financial one. If courts treat safety A/B tests as evidence of design defect, then every platform's experimentation playbook becomes a discovery target. The cost is not the settlement alone; it is the change in how product teams are allowed to test. A feature that could reduce harm but also reduce time-spent is no longer just a product tradeoff. It is a documented decision that a jury may weigh against the company.
The Legislative Clock: Two Bills, One Pressure Point
The timing of the letter is not accidental. On August 5, the Senate Commerce Committee unanimously advanced the Kids Online Safety Act. The Senate bill includes a "duty of care" section requiring covered platforms to "exercise reasonable care in the creation and implementation of any design feature to prevent and mitigate" specified harms to minors, including eating disorders, substance use disorders and suicidal behaviors. The House passed its own version, the Kids Internet and Digital Safety Act, on June 29 by a 267-117 vote, bundling a revised KOSA with an updated Children and Teens' Online Privacy Protection Act.
The two chambers are now throwing down competing gauntlets. Senate Republicans and Democrats united behind a design-focused, content-neutral standard; the House package leans harder on privacy and enforcement architecture. Blumenthal and Blackburn called the House bill "dead on arrival" after it passed. The practical effect is the same either way: for the first time, a federal duty of care for child safety is within striking distance of becoming law, and the TikTok document gives enforcers a concrete example of what "unreasonable" design looks like.
Across the Atlantic, the pressure is already operational. The European Commission issued preliminary findings in July that TikTok failed to protect children's privacy under the Digital Services Act, allowing minors' accounts and content to remain too widely visible to adults and people without TikTok accounts. The US and EU tracks are converging on the same point: safety claims must be demonstrable, not asserted.
The Second-Order Question: What Gets Tested After This
The first-order consequence is reputational and legal for TikTok. The second-order consequence is structural for the entire industry. Platforms run A/B tests on virtually every product change — ranking, notifications, autoplay, friction. The TikTok document shows that safety features were tested the same way as any other feature, with engagement as the readout. If that becomes legally actionable, companies face a new asymmetry: testing a safety feature and finding it reduces engagement creates evidence; not testing it leaves the company blind to whether safety and growth actually conflict.
There is a real tradeoff buried in the numbers. TikTok's own document estimated that even its proposed fix — a 1% control group — would still leave at least 1.8 million users without the safeguard. The company was measuring a cost it had already decided to impose at some scale. The question for investors is how much of the engagement model depends on exactly these kinds of marginal optimizations. If the industry-wide answer is "a lot," then the regulatory endgame is not a fine. It is a redesign of the revenue engine.
The Counter-Thesis: Testing Is Standard, and Disclosure Is the Real Failure
The strongest case against the senators' framing is that A/B testing is an industry norm, not a TikTok aberration, and that the moral failure here is not the test but the silence around it. Former employees and industry practitioners note that backtests are a routine way to validate that a change does not break the product. From this view, the "depraved" label conflates a methodological choice with a malicious one. The company did, after all, write the review, recommend shrinking the control group, and shorten the test window.
That defense has limits. Routine testing does not require withholding a safeguard from 15 million users for up to a year. Routine testing does not exclude an entire class of vulnerable users by default. And routine testing does not omit the one known harm case when answering Congress under questioning. The counter-thesis survives only if the test was short, small, and disclosed. On the record so far, it was none of the three.
The falsifying signal for the regulatory thesis is specific: if the House and Senate bills fail to reconcile before the end of the 119th Congress, or if courts grant platforms broad immunity for experimentation conducted under documented safety protocols, then the liability overhang is priced as noise rather than regime change. Conversely, if KOSA clears both chambers with the duty-of-care language intact, the TikTok document becomes the exhibit that defines the standard.
What Comes Next: Exposure, Beneficiaries, and Scenarios
The near-term exposure is concentrated in the platforms with the largest teen user bases and the most aggressive engagement optimization: TikTok, Meta's Instagram, Google's YouTube, and Snap. All four are already defendants in the teen-mental-health litigation. The TikTok document raises the discovery risk for each of them, because the same experimentation infrastructure exists across the sector.
The medium-term beneficiaries are the companies selling the alternative: age-assurance vendors, safety-tooling providers, and platforms that can credibly claim a duty-of-care posture as a product feature. The Kids Online Safety Act would create a compliance market almost overnight. The long-term structural question is whether the attention-economy model can absorb a binding constraint on engagement optimization without giving up growth.
Base case: KOSA or a close variant passes the Senate but stalls again in the House, leaving the litigation track to set the standard through settlements and occasional verdicts. Upside case for platforms: the House bill's narrower framework wins, limiting the duty of care and preserving experimentation latitude. Downside case: a federal duty of care becomes law with the Senate language, and the TikTok document is entered as evidence in the bellwether trials, establishing that safety A/B tests are admissible proof of design defect.
Short term, expect more letters, more subpoenas, and more settlements before trial. Medium term, expect product teams to narrow or restructure safety experiments and to document consent and exclusion criteria as if they will be read by a jury — because they will. Long term, the industry either proves that safety and engagement can be optimized together, or it accepts a lower growth ceiling as the cost of a social license it has been unable to earn.
The central judgment: this is not a moderation scandal. It is a measurement scandal. TikTok did not merely fail to protect a child; it instrumented the failure to see what it cost the business. That distinction is what makes the document dangerous, and it is what will make the liability stick.
Explore more exclusive insights at nextfin.ai.
