Azernews.Az

by Alimat Aliyeva

TikTok conducted experiments in which thousands of users, including children and teenagers, were reportedly given non-working versions of certain safety features.

The allegations concern so-called “placebo experiments,” in which some users received a broken version of TikTok’s Algo Refresh feature. The tool is designed to reset a user’s recommendation algorithm and remove unwanted or potentially harmful content from their feed.

Users who were included in the experiments believed they had successfully activated the feature. However, their recommendation feeds did not actually change. Another group of users received a working version of the tool, allowing TikTok to compare the behavior of people whose feeds had been genuinely reset with those who believed they had been.

According to the lawsuit, TikTok repeatedly conducted such experiments in 2023, involving thousands of users. The company reportedly compared how long people stayed on the platform, how they interacted with content, and how the changes affected advertising revenue.

The experiments reportedly raised concerns even inside TikTok. One manager responsible for users’ digital well-being warned colleagues that the tests could mislead users.

“This feature is designed to ensure algorithmic transparency and user control. The placebo group completely contradicts that,” the employee wrote in an internal chat, according to the lawsuit.

New York authorities also accuse TikTok of misleading users about how Algo Refresh actually works. According to the lawsuit, the company presented the feature as a way to reset the recommendation algorithm, while the effects were allegedly only temporary.

A TikTok representative told Reuters that the allegations misrepresent how Algo Refresh works. The company said that, like other technology companies, it regularly tests new products and features to evaluate user experience and understand how they perform in real-world conditions.

The released court materials also describe another series of TikTok experiments, previously reported by Bloomberg. In those tests, the company allegedly selected users at random and disabled certain safety features designed to restrict access to potentially harmful content.

The most serious allegations involve teenagers. According to Bloomberg, one New York teenager who participated in the experiment later died by suicide after being exposed to content related to self-harm and suicide. The lawsuit claims that an internal TikTok audit later found a significant amount of such material in the teenager’s feed.

New York Attorney General Letitia James filed the lawsuit against TikTok in 2024. New York was among more than two dozen U.S. states that accused the company of contributing to addictive behavior among children and misleading users about the safety of its platform.

TikTok has rejected the allegations, maintaining that the safety and well-being of younger users remain among its priorities.

In September 2026, TikTok settled the first of the lawsuits brought by U.S. states that was expected to go to trial. The company agreed to pay Alabama up to $100 million and make changes to how its platform operates for teenagers in the state.

Earlier, TikTok and the U.S. Department of Justice reached an agreement over alleged violations of the Children’s Online Privacy Protection Act (COPPA). TikTok and its parent company, ByteDance, agreed to pay $400 million.

The controversy raises a broader question about how far technology companies should be allowed to go when experimenting with their users. A/B testing is a normal part of developing digital products, but the situation becomes much more complicated when experiments involve children, mental health, potentially harmful content, and features that are specifically presented as tools for user protection.

It also highlights a fundamental problem with recommendation algorithms: users may believe they have control over what appears in their feeds, while the systems making those decisions remain largely invisible. If a safety feature can appear to work while quietly doing nothing, users may have no way of knowing whether they are actually protected.

That is what makes the TikTok allegations particularly significant. The issue is not simply whether an algorithm works as intended — it is whether users can trust the platform when it tells them that a safety feature is protecting them.