A hot potato: In 2020, Meta launched an internal research effort called Project Mercury to examine the psychological effects of Facebook use. Working with market research firm Nielsen, the company ran a survey-based experiment that asked selected users to deactivate their Facebook accounts for a week. Newly unsealed discovery documents reveal that those who stepped away reported noticeably lower levels of depression, anxiety, loneliness, and negative social comparison.
Instead of publishing the findings or commissioning further research, Meta's leadership shut the project down. Internal correspondence, cited in a lawsuit brought by US school districts, reveals that the company questioned the study's validity, attributing its results to prevailing negative press about social media.
At least one researcher involved in the project argued internally that the survey pointed to a causal link between Facebook use and social comparison stress. Internal discussions even likened Meta's handling of the data to the tobacco industry's efforts to obscure the health effects of cigarette use.
According to documents cited in the complaint, Meta also failed to share these results with Congress, maintaining instead that it had no definitive evidence of harm to teenage users, especially girls – a position that conflicted with its own internal research.
The filing includes detailed descriptions of Meta's product design and safety-policy decisions, outlining how many youth safety tools were engineered in ways that made them rarely used.
The filing goes beyond the buried study. It includes detailed technical descriptions of Meta's product design and safety-policy decisions, outlining how many youth safety tools were engineered in ways that made them rarely used. Internal tests of more restrictive features were often blocked over concerns they might hurt user-growth metrics.
Meta is further accused of keeping a high "strike threshold" for serious offenses like sexual exploitation, requiring users to be caught more than a dozen times before being removed from the platform.
The documents also describe how engagement-optimization algorithms designed for teens boosted the spread of content with harmful psychological effects, including material tied to body dysmorphia and eating disorders. Safety engineers' attempts to reduce these risks were deprioritized, the filing claims, with leadership framing inaction as a business necessity.
Beyond formal safety tools, internal emails show Meta executives discussing how increasing privacy for teenage live video streams could help retain youth engagement, even if it reduced oversight from parents or educators.
Meta spokesperson Andy Stone has rejected the allegation that the company intentionally concealed negative findings, saying the Facebook deactivation study was abandoned due to methodological flaws, not because of its conclusions.
Stone told Reuters that Meta has spent years developing parental controls and teen safeguards, describing the resulting safety tech as "broadly effective." Meta argues that the documents referenced in the complaint are being taken out of context, and that its motion to strike them relates to privacy and redaction procedures – not to suppressing evidence.
The plaintiffs, meanwhile, accuse Meta (along with Google, Snap, and TikTok) of systematically downplaying known risks to young users. The filings also reference internal communications suggesting that several companies sponsored third-party child advocacy groups at the same time those groups publicly endorsed the platforms' safety practices.
The information released in the Northern California federal court filing is part of a much larger multidistrict litigation effort. Plaintiffs allege that major platforms not only built addictive, engagement-driven algorithms but also delivered less robust safety protections to U.S. teens than to children in other countries.
Internal presentations at TikTok, for example, described offering "spinach" – healthier, more educational content – to children on its Chinese app, Douyin, while American children received "opium," a label for hyper-engaging content that could be harmful. TikTok disputes both the accuracy and characterization of the comments.
Snapchat's internal emails acknowledged that features like "Snapstreaks," which encourage daily communication, effectively cultivated compulsive behavior among teens.
Even as legislative pressure mounts, the defendants have continued to argue that their algorithmic feeds and youth-targeted features are designed with safety in mind, and that their policies draw on guidance from external experts. Major platforms, including Meta and Google, are currently challenging new California rules that would limit algorithmic recommendations to minors without explicit parental consent.
The outcome of these cases could set major precedents for how internal research is disclosed and regulated across the tech industry. With the next hearing set for January 26 in the US District Court for the Northern District of California, regulators, researchers, and industry watchers are looking closely at how much of Silicon Valley's internal decision-making may soon be forced into the open.
Meta shut down an internal study that linked Facebook use to worse mental health
