What the complaint alleges
A New York complaint against TikTok, unsealed on Wednesday 7 October, alleges the company ran experiments on thousands of users who tried to switch on one of its safety features and handed some of them a version that did not work.
The feature is Algo Refresh, a control that is supposed to reset a user’s recommendations so they can get out of a stream of unwanted or potentially harmful content. According to the complaint, reported by Reuters, the users given the non-working version believed the reset had happened. Their feeds did not change. Users in the group that received the working version saw their feeds change.
The complaint says the tests were run repeatedly in 2023, and that the two groups were compared on how long people spent on the site and on the effect on advertising revenue. The children and teenagers who opted into the feature were among those tested.
The internal objection
A product manager on TikTok’s Digital Well Being team objected in an internal chat quoted in the filing. The feature, the manager wrote, “is about algorithm transparency and control” — and “a placebo group completely conflicts with that.”
The complaint makes a second claim about the same control: that TikTok told users Algo Refresh would reset their recommendation algorithm when the change was in fact only temporary.
A separate allegation in the lawsuit, previously reported by Bloomberg, is that TikTok randomly turned off safety features limiting exposure to harmful content for some users. The complaint says a teenage boy in New York who was in that test group died by suicide, and that internal reviews showed he had been shown a large volume of self-harm and suicide content.
Where the case stands
Attorney General Letitia James sued TikTok in 2024. These claims were added in an updated version of the complaint filed at the end of August and kept under seal until a judge ordered it opened. The case is one of more than two dozen brought by US states accusing the platform of designing a service to be addictive for children and of misleading users about its safety tools.
In September TikTok settled the first state case that was due to go to trial, agreeing to pay Alabama up to $300 million and to change how the platform works for teenagers in that state.
TikTok has broadly denied New York’s allegations in court filings and says keeping young users safe is a priority. A spokesperson told Reuters the company “continuously make[s] improvements” and, like most companies, regularly tests features to understand how they work in the real world. The spokesperson said the allegations mischaracterise how Algo Refresh works.
Why this matters beyond TikTok
Everything here is an allegation in a civil complaint, not a finding. But the thing being alleged is specific and it is not unique to one company: that a control offered to users as a way to steer a recommender was itself the subject of an experiment, with the outcome measured in session length and ad revenue.
Platforms A/B test almost everything, and a safety control is a feature like any other in that pipeline. The complaint’s argument is that it is not — that once a company tells a user a switch resets the algorithm, shipping a switch that does nothing is a statement about the product rather than a measurement of it.
What to watch
Whether the court lets the placebo claims proceed, and whether any of the other state cases adopt the same theory. The Alabama settlement showed what a trial date is worth; a second one with this allegation attached would be worth more.
