Kids Over Clicks

It's time to keep kids as safe online as we do on the road or at home.

The typical American teen spends about five hours a day on social media and nearly one in three kids demonstrates an addictive pattern of social media use by the time they finish middle school — one researchers have linked to higher rates of anxiety, depression, and suicidal ideation. At the same time, they have grown lonelier, while reading and math scores have fallen sharply. For generations, government has stepped in to keep addictive products like tobacco and alcohol out of young people’s hands. And from cribs to car seats to bike helmets, the physical products in a kid’s life have to pass federal safety tests. It’s time we hold the digital products in their lives to the same standard and implement a comprehensive online safety agenda that puts kids over clicks. 

The Problem:

In 2013, more than half of American high-school girls — 53 percent, up by more than a third from a decade earlier — reported persistent feelings of sadness or hopelessness, which the CDC has linked to the saturation of smartphones and social media use in adolescent life. Adolescents who use social media more than three hours a day are roughly twice as likely to experience symptoms of depression and anxiety as non-users, and heavy adolescent social media use is correlated with measurable declines in cognitive performance. College professors report that students now struggle to read continuous prose for more than a few pages without reaching for the phone. Former U.S. Surgeon General Vivek Murthy, citing this impact on mental health, called for a tobacco-style health warning on social media.

An entire generation of kids has grown up with unfettered access to social media platforms. As time goes on, the impact has become clear. At its best, social media can foster connection and communication. But study after study has also shown that heavy social media use during adolescence hijacks young people’s attention, reduces their ability to concentrate on tasks and retain information, and puts them at higher risk for anxiety and depression.

We are witnessing the “tobacco moment” for social media and artificial intelligence (AI) giants. In the 1990s, parents, states, courts, and the federal government came together to prevent cigarette companies from targeting kids. That same groundswell of opposition is now mobilizing against social media. At least 42 states, hundreds of school districts, and more than a thousand individual plaintiffs have filed lawsuits alleging that these platforms knowingly deployed addictive design features that, in many cases, led to depression, eating disorders, self-harm, and suicide. In a landmark decision, the verdict in the first of these trials ruled against Meta and Google, finding that they harmed a young user with design features that were addictive and contributed to anxiety, depression, and other mental health struggles.

The stories in these lawsuits echo the case of Molly Russell, a 14-year-old who died by suicide in 2017 after platforms repeatedly pushed self-harm and suicide content into her feed unsolicited. A formal investigation found the material contributed to her death “in a more than minimal way.”

Too little has changed. In 2025, the foundation set up by Molly's family ran an experiment using dummy accounts posing as a 15-year-old girl. The result: 97% of recommended videos on Instagram Reels and 96% on TikTok were flagged as harmful, and 55% of harmful posts on TikTok's For You feed referenced suicide and self-harm.

The pattern is now repeating with AI chatbots. In 2024, 14-year-old Sewell Setzer III died by suicide after months of intimate, sometimes sexual exchanges with a chatbot modeled on a Game of Thrones character. In his final conversation, he promised the bot he would "come home" to her. It replied: "please do, my sweet king." Moments later, Sewell died by suicide.

What follows is a sweeping vision for protecting kids online. This paper focuses on kids and teens, whose developmental stage, exposure to harm, and limited capacity to consent demand heightened protection. The proposals that follow are centered on how they would benefit kids. But many of these reforms — including strong privacy settings by default, strict data minimization standards, and a ban on surveillance advertising — would benefit adults too, and we see no reason why these protections shouldn’t be applied to all users.

Technology has transformed American life — mostly for the better — and the next wave of innovation, including artificial intelligence, holds even more promise. But that promise will be wasted if the next generation of kids grows up anxious, addicted, and unsafe. To date, the federal government has largely sat on the sidelines when it comes to establishing commonsense rules of the road. The next President has the opportunity to change that. 

The Solution:

Age Restrictions

Lawmakers should ban social media platforms from opening accounts for children up to age 16. By keeping the youngest users off these platforms entirely, this approach buys time for adolescent brains to develop the cognitive defenses needed to navigate platforms built around addictive design. These restrictions are similar to a Florida law that received overwhelming bipartisan support and similar measures that have already been implemented in Australia. Many other countries have also taken steps toward implementing age restrictions for social media, as has the European Union. 

Age restrictions have been an important public policy tool to protect minors from dangerous but legal conduct in many other areas (tobacco, driving, gambling, etc). This policy is not a silver bullet. Yes, some kids will use VPNs and other mechanisms to get around the ban, just as some kids buy fake IDs to buy beer at a liquor store. And any age restrictions must be carefully tailored to balance legitimate privacy and civil liberties concerns, as New York Attorney General Letitia James recently demonstrated in her proposed rulemaking to implement the Stop Addictive Feeds Exploitation (SAFE) for Kids Act.

Privacy-by-Default and Safety-by-Design

Lawmakers should provide privacy-by-default and safety-by-design protections, drawing on models already in place in Maryland, Nebraska, Vermont, and South Carolina. These protections are common sense: opening new accounts with the highest privacy settings by default, restricting addictive features, banning manipulative design tactics, enforcing strict data minimization and privacy practices, disabling private messaging from strangers unless a user knowingly opts in and accepts them, providing for timely account deletion, and publishing clear policies for users to understand their rights and options. 

These privacy-by-default settings work to close a crucial threat vector through which predators target kids. According to internal documents from Meta's own child safety researchers, roughly 500,000 children per day are targeted by sexually exploitative messages on Instagram and Facebook in English-language markets alone — and the researchers acknowledge that the true number is almost certainly higher. An adult predator can't send an unwanted sexual message to a child whose account doesn't accept messages from strangers. They can't search for and find a child’s account through a "People You May Know" recommendation if the child's account doesn't appear in those recommendations to begin with. And they can't lift photos or identify a school name or a hometown off a profile that doesn't display them. Banning design features like infinite scroll, autoplay, “Streaks,” and surgically timed push notifications in accounts for children switches off the machinery that leads to compulsive use. 

Each of these heightened default settings — disabling messages from strangers, turning off addictive features, making account deletion easy — is a small structural change on its own. Together, they redesign the product around a kid’s wellbeing and safety rather than around the platform’s engagement metrics.

Rules of the Road for AI Chatbots

Generative AI chatbots — consumer-facing AI systems that can hold conversations, role-play characters, and answer the kinds of questions previously brought to search engines — are fundamentally new and our laws have not caught up. Other industries that pose this kind of risk to the public have rules of the road: cars are crash-tested, drugs are screened for safety and efficacy, and health professionals like doctors and therapists are licensed. Chatbots have none of that. The regulatory environment is slipping further and further behind and lawmakers need to act quickly to put bright-line rules in place before more harm is done. 

A critical first step is clarifying in statute that technology’s special liability shield, enshrined in Section 230 of the Communications Decency Act, does not immunize companies from accountability based on the outputs of their AI systems. Even Senator Ron Wyden, a co-author of Section 230, agrees that Section 230’s protections do not apply to content generated by AI. This clarification would allow kids who experience chatbot-induced harms to have their day in court and force companies to internalize some of the downstream costs, leading to incentives more aligned with the public interest.

In addition, lawmakers should draw bright-line rules around “therapy chatbots” and prohibit any chatbots from representing themselves or their outputs as provided by a licensed professional such as a lawyer, doctor, or psychiatrist. A licensed professional has numerous obligations to their patient or client, including confidentiality, a duty of care, mandatory reporting, and malpractice exposure. A chatbot cosplaying as a licensed professional has none of those obligations and should not be able to benefit from the trust carried by licensed professionals. 

Lawmakers should also establish baseline privacy protections that limit AI companies from using chat histories for advertising or any other business purpose in the same way doctors, lawyers, and therapists have legal, ethical, and professional obligations to protect their patients’ and clients’ information. These protections matter most for kids. The data file a chatbot builds on a teenager is more intimate than anything a social media app has ever built — not what a child clicked on or scrolled past, but what they said in their own words when they thought no one was listening. A teenager who confides in a chatbot about their depression, body image, or what is happening at home should not, an hour later, be served weight-loss ads or content engineered to keep them in that negative space for as long as possible because that’s what’s best for advertisers.

Due to the extensive harm that can result when unsafe AI products are rushed to market, lawmakers should impose strict liability for developers of chatbots that cause demonstrable harm to users, as seen in the cases discussed above. Had these bright-line rules been in place in January 2026, Elon Musk’s Grok chatbot would have faced meaningful accountability for the thousands of nonconsensual sexual deepfakes it produced, including — by one estimate — 23,000 sexualized images of children.

Bell-to-Bell Cellphone Bans

American students' reading and math scores on the National Assessment of Educational Progress have fallen to historic lows, with the steepest drops among the cohort that has had a smartphone since middle school. College professors find themselves shortening class readings to adjust to students’ fleeting attention spans. And the available causal evidence — most of it from schools that banned phones before their neighbors did — points in one direction: when the phones go away, students’ learning experience improves.

Lawmakers should create a national standard for bell-to-bell cellphone bans in schools, building on laws that 35 states and D.C. have passed to limit cellphone use in the classroom. While many of these laws are just beginning to take effect, early observations suggest that removing cellphones from the classroom can lead to improved academic outcomes and richer social interaction between classes. One year into a cellphone ban, a school district in Dallas, Texas, saw a nearly 25 percent increase in library books being checked out.

A properly implemented bell-to-bell ban needs three sets of carve-outs. First, accommodations for those with medical needs, emergencies, and individualized education plans. Second, for student caregivers who are responsible for the health and safety of their kids, elderly relatives, or younger siblings. And third, for non-internet-connected phones that should be kept off and away during the day but preserve the ability for kids to call or text in the event of a genuine emergency such as a school shooting. 

Launch a “Free to be Kids” Campaign — A Smartphone-Free Childhood Until Age 14

Ask parents of a tween whether they want their kid to have a smartphone, and most will say no. Ask whether their kid has one anyway, and most will say yes. It’s a collective action problem: once one kid in a friend group or class gets a smartphone, the rest follow. No individual family can fix this phenomena alone. 

New law isn’t needed; rather, the next Administration can use already-available tools to empower parents on something most families already want but cannot enforce alone. The Surgeon General could issue formal guidance on screen time and smartphone use and the federal government should begin publishing regular national data on children's smartphone use, social media use, and time spent with AI chatbots, so the American public can see what is happening to children's attention as clearly as it sees the unemployment rate or the inflation index. 

The President should issue a primetime address on the topic, and the White House could lead coordinated outreach to trusted voices like faith leaders, youth-serving organizations, and the country's largest school systems. If enough schools, districts, and communities sign on publicly, no kid has to be the odd one out, and no parent has to absorb the social cost of being the only family in the friend group to say no.

Dismantle the Attention Economy

The attention economy is fundamentally built on surveillance. The platforms that demand kids’ attention make money by tracking their activity — on their own platforms and beyond — amassing troves of personal data to sell or share with advertisers or data brokers. These digital dossiers include every click, view, purchase, keystroke, and biometrics, including eyeball movement on a screen. Every American kid who has ever used a social media platform or owned a smartphone has this kind of dossier, held by companies they have never heard of, traded by data brokers with no way to opt out, and used for purposes without their knowledge or consent.

Many of the gravest threats to kids’ online safety are byproducts of the attention economy. These individualized records, endlessly expanded by the attention economy, are what make a teenager struggling with body image identifiable to an advertiser at the exact moment they are most likely to click on a weight-loss ad. They let a platform’s “People You May Know” algorithm surface a teenage girl to a predator who has shown an interest in teenage girls. They feed the recommendation algorithms that take a kid’s single curious search about eating disorders and turn it into a rabbit hole of pro-anorexia content. Without this data infrastructure, none of these harms would be commercially profitable or rational — and most of them would not happen at all. It is not possible to fully fix kids’ online safety without addressing the data engine that powers the platforms harming them.

The next President and Congress can dismantle the attention economy and strengthen data privacy through two major steps: 

First, implement strict limits on kids’ data collection itself, not just on data use. This approach, what privacy experts call “data minimization,” provides that companies are not allowed to collect or hold personal data they do not actually need to deliver the service the user signed up for and should include a flat ban on the sale of precise geolocation data. Applied to AI chatbots, data minimization means preventing AI companies from training models on personal data without affirmative consent. Policymakers are already beginning to apply these protections broadly; in Maryland, for example, recent legislation that passed with overwhelming bipartisan support imposes strict minimization requirements for all accountholders. Further, every American should have an easy-to-follow path to delete data that companies and data brokers have already collected about them. California residents now have this right through the California Privacy Protection Agency’s DROP System implemented earlier this year. 

Second, ban surveillance advertising to remove the single biggest financial incentive for companies and data brokers to invade children’s privacy. In practice, this kind of advertising allows a teenage girl who has just deleted a selfie to be served a beauty product ad within the hour, because a platform’s data has flagged her as emotionally vulnerable and therefore likely to buy. It is not about banning ads altogether. Online platforms would still be able to run contextual ads — ads served to users based on the content they are viewing, for example, an ad for makeup near a “get ready with me” influencer’s video — and even geo-target ads to a kid’s broad location (e.g, municipality) so that children in Miami, Florida aren’t seeing ads for a bakery based in Cleveland, Ohio. Online platforms could continue generating billions of dollars from selling ads; they just wouldn’t be able to sell ads based on a child’s personal data or current emotional state. 

Make Platforms Pay for Harm They Cause Kids

For decades, technology companies have been insulated from the legal and economic consequences of their business practices — a result of Section 230, which provides broad immunity to online platforms for user-generated content. But the current digital marketplace was unimaginable when the law was written. Section 230 protections should be narrowed in four specific ways. 

First, to exclude paid ads. Investigations from the Tech Transparency Project (TTP) demonstrate how readily social media platforms push dangerous content on young people’s feeds. In 2024, for example, TTP used Meta’s services to produce and place ads targeting users ages 13-17 promoting “pill parties,” extreme weight loss, and teen vaping. Meta’s ad-approval system cleared every one of these ads in less than five minutes. Second, to exclude illegal content or activity (e.g, illegal drug sales, sex trafficking) once the platform has knowledge of it. Third, to exclude platforms that algorithmically promote content that constitutes stalking, harassment, or nonconsensual intimate imagery. And fourth, to exclude any AI service that creates or co-creates content with a user, such as chatbots, image generators, or video generators. 

These reforms would align platforms’ incentives more fully with promoting the health, safety, and welfare of everyone, but girls and women in particular, who are disproportionately the victims of online harms. 

Conclusion:

The choice that some pose between optimism about technology on the one hand and protecting kids and the values of the American people on the other is a false one. In every generation, American leaders have shown that smart rules of the road for new technologies promote innovation while keeping Americans safe. Today, the harms many children experience online undermine trust in the entire technology sector — making it harder for the lawmakers who want to champion the innovation economy America needs.

A comprehensive digital policy agenda can keep our kids safe from digital addiction and other online harms while creating an environment that allows them to experience the incredible opportunities for connection and learning inherent in new technologies. More fundamentally, it can restore the basic premise that technology should serve people and not the other way around. The policy agenda outlined in this paper provides a pathway to make that environment real. 

By: Rishi Bharwani, US Director of Reset Tech