Countries around the world have taken significant steps in the last year to treat children differently than adults on the internet. Australia’s social-media age limit for users under 16, for example, went into effect in December; the United Kingdom passed a similar measure this summer. New Zealand and other nations are weighing age restrictions. Here in the United States, in a bellwether trial against major social-media companies this past spring, a Los Angeles jury ordered Meta and YouTube to pay millions to a young victim due to the addictive nature of their platforms.
Parents have been told for decades that the duty of protecting children online is theirs alone. But they are no longer willing to accept this arrangement — it simply isn’t working. State legislators are responding to their constituents’ demands with parental-consent requirements, design specifications, and even outright bans. The question facing those lawmakers, and the courts reviewing their work, is how to regulate these technologies in a manner that accords with the Constitution and fits into America’s legal tradition.
Our legal system has developed three methods of regulating children’s engagement with certain harmful products and activities: prohibition, parental consent, and unrestricted access. It’s worth looking more closely at these categories to determine how they might be applied to the modern digital ecosystem, beginning with prohibition.
PROHIBITION
As a society, we have decided that some products and activities should be off limits to minors, and no parental signature can change that. One prominent example is tobacco. During the 19th century, children as young as five could buy and smoke cigarettes; by 1939, all states had age restrictions on their purchase and use. These restrictions came well before definitive medical evidence of cigarettes’ lethality emerged; the intuition that smoking was bad for children preceded the science by decades.
Today, federal law prohibits the sale of tobacco products to anyone under 21, and almost all states require age verification for such sales. The reasoning is straightforward. Tobacco causes addiction and long-term negative health consequences that children cannot meaningfully appreciate. We don’t trust children to make decisions regarding trade-offs across such lengthy time horizons, and we don’t rely on parents to make these choices for them, either. Some risks are just too great.
Bans on minors’ purchase and consumption of alcohol are grounded in similar reasoning. As in the case of tobacco, alcohol use is associated with serious, long-term harms like addiction and developmental problems. But it also comes with more immediate risks, among which are impaired judgment, drunk driving, and alcohol poisoning.
Before 1984, states determined their own age restrictions on alcohol. That year, however, Congress passed the National Minimum Drinking Age Act, which threatened to withhold federal highway funds from states that didn’t prohibit individuals under 21 from purchasing and publicly possessing alcohol. By 1988, all 50 states had complied. Today, every state prohibits selling alcohol to those under 21, and the vast majority require age verification for purchase.
Gambling offers another clear example of an age-based categorical prohibition. Gambling is a concern for minors because children can’t assess odds rationally, and early exposure to the activity increases one’s risk of developing a gambling addiction, which can be devastating. In America today, states set their own age minimums for different forms of gambling, typically ranging from 18 to 21. Children and youths under 21 are not legally permitted to buy lottery tickets or place sports bets. Some states even prohibit them from being present in casinos and other facilities where gambling is taking place — a recognition of the dangers associated with merely exposing children to a highly addictive activity.
Firearms present a fourth example of a categorical ban. Again, prohibition makes sense here because improper use of a firearm can have devastating and irreversible consequences. In 1968, Congress established the first federal age minimums for purchasing guns from licensed dealers — 18 for long guns and 21 for handguns. The Youth Handgun Safety Act of 1994 went further, banning possession of handguns by people under 18 and prohibiting adults from transferring handguns to juveniles. Today, many states impose additional restrictions on minors’ purchase and use of firearms.
The Constitution adds an interesting wrinkle to firearms restrictions, as the Second Amendment protects Americans’ right to bear arms. But even the most expansive reading of that right has accommodated age-based limits. In District of Columbia v. Heller, the Court left undisturbed “laws imposing conditions and qualifications on the commercial sale of arms,” the category into which age restrictions on gun sales have traditionally fallen. What courts are debating now sits at the edge of adulthood, not childhood. After New York State Rifle & Pistol Association v. Bruen reoriented Second Amendment analysis around historical tradition, the Fifth Circuit, in Reese v. Bureau of Alcohol, Tobacco, Firearms, and Explosives, struck down the federal ban on handgun sales to 18-to-20-year-olds, while the Fourth and 11th Circuits upheld comparable under-21 limits in McCoy v. Bureau of Alcohol, Tobacco, Firearms, and Explosives and National Rifle Association v. Glass. No one suggests a state cannot bar minors from buying firearms; the only open question is how far that authority reaches into young adulthood.
These examples suggest that two factors render age restrictions on an activity or substance appropriate. The first is that it presents serious risks of long-term, often irreversible consequences, as well as a high potential for addiction. Tobacco, alcohol, gambling, and firearms all pose significant health or safety dangers to children, harms that may not materialize for years or even decades. The first three of these are also highly addictive — especially for children and adolescents, whose brains are still developing.
The second factor is that the activity or substance is associated with risks that exceed the capacity of household-level parental decision-making to manage. Categorical bans tend to emerge when parents themselves conclude they cannot effectively handle the danger family by family. The campaign to raise the drinking age to 21, for instance, was powered by parent advocacy. Mothers Against Drunk Driving and allied groups argued that individual families could not counteract the social pressures and negative group dynamics of a permissive legal environment for alcohol. In this case, a socially based threat required lawmakers to step in and assist parents in protecting their children.
PARENTAL CONSENT
Whereas prohibition supports parents by taking the decision out of their hands, the second category of restrictions ensures that parents are given a choice. Parental-consent laws restrict minors’ access to activities that involve possible risks but also offer them potential benefits. The law allows children to partake in them, but only when parents have evaluated the risks and authorized their child to participate. In many cases, activities in this category follow a hybrid model wherein the law imposes certain minimum-age restrictions or other limits, but in addition to those limits, parents must provide consent for their child’s participation.
Employment of minors is a classic example of a consent-based restriction. Remunerative work offers teenagers important benefits, including economic independence and greater financial literacy. It can also help them develop work ethic, time-management capabilities, and professional skills. Research on adolescent employment, however, shows that work can also harm minors. Moderate work hours (under 20 per week), for example, correlate with positive outcomes, while excessive time on the job can interfere with school performance and increase a teen’s tendency to engage in risky behaviors.
These mixed effects lend support to a hybrid legal regime whereby the state imposes age minimums, hour restrictions, and the like to protect children from exploitation and developmental harm — but once those thresholds are met, parents must consent to their child’s employment based on his circumstances and maturity. Approximately 38 states mandate employment certificates or work permits for minors, often incorporating parents’ signatures into the application process.
Participation in youth sports also relies on parental consent. Joining organized athletics teams offers a wide range of benefits for children. It teaches them self-discipline, requires them to engage in physical activity, helps them develop leadership and teamwork skills, gives them a chance to socialize with other children, and more. But playing sports also carries risks, ranging from minor injuries to potentially catastrophic harm. Consent forms and liability waivers ensure that parents are aware of the risks and have authorized their child to participate. They also encourage parents to take precautions to reduce the chances of their child’s being injured.
Tattoos and body modifications are another hybrid area of the law. Although tattoos are considered a form of protected expression under the First Amendment, states are allowed to prohibit children from obtaining them or require parental consent for minors to receive them once they reach a certain age. Today, all 50 states require individuals to be at least 18 years old to receive tattoos — a limit based on the legal principle that minors cannot provide informed consent for permanent body modifications. Most states, however, also permit minors ages 16 or 17 (and sometimes as young as 14) to receive tattoos with written parental consent. Many of these states require both the child and a parent to appear in person with government-issued identification.
This hybrid model reflects the recognition that body modifications cause long-lasting physical changes that require maturity to make, while also acknowledging that parents can adequately assess whether their teen, having reached a certain age, is mature enough to make decisions on such matters. It also ensures that parents can oversee the healing process.
Driving presents a similar case. All states set a minimum age for learner’s permits (typically 14-16) and full licenses (16-18), recognizing that younger teenagers lack the judgment, reaction timing, and risk-assessment skills necessary to operate motor vehicles safely. Within these age boundaries, the vast majority of states also require parental consent for minors to obtain learner’s permits and driver’s licenses.
The hybrid framework of parental authority with state-imposed age restrictions as well as graduated licensing requirements reflects a nuanced judgment about driving. While it offers significant developmental benefits — including greater independence, access to employment, a chance to engage in more social and extracurricular activities, and an opportunity to develop practical life skills — it also carries substantial risks, as motor-vehicle crashes remain the leading cause of death for teenagers in the United States. Parental involvement thus serves multiple functions: Parents can gauge their own child’s readiness, provide supervised opportunities for their child to practice driving, and monitor their child’s compliance with laws and other restrictions. In many states, parents bear responsibility for any damage that their child’s negligent driving may cause.
All these activities in the parental-consent category share several characteristics. First, the activity often carries benefits alongside its risks, and parental involvement can help reduce the harms involved. In many cases, the dangers vary substantially depending on a minor’s circumstances and maturity, making parental consent a useful screen that leverages parents’ knowledge of their own children when determining whether they should be able to participate. And often the value of obtaining parental consent lies less with initial approval than in parents’ ongoing involvement, whereby parental supervision actively lessens the risk. This differs from activities subject to categorical bans, where no amount of oversight will make them safe.
Second, harms associated with activities in the parental-consent category are often limited in degree and time. Sports injuries may be devastating when they occur, but the vast majority of them will heal, leaving no significant permanent damage. A tattoo, though considered permanent, is technically removable (albeit at great expense and additional pain). The law tolerates some suboptimal parenting as the price of family sovereignty, as long as the risks aren’t long term, catastrophic, or irreversible.
Other activities fall into the parental-consent category because they have long been understood by tradition and cultural norms as belonging in the family’s domain. The Supreme Court has recognized and protected parental authority in cases like Meyer v. Nebraska, Pierce v. Society of Sisters, and Troxel v. Granville, all of which held that the American legal tradition protects parents’ rights to oversee their children’s education and upbringing.
One important corollary here is that parents’ fundamental right to direct the raising of their children applies not only against the government, but also against third parties. The Supreme Court has recognized as much in cases like Rowan v. U.S. Post Office Department and Federal Communications Commission v. Pacifica Foundation. This is an important principle of the law to consider when regulating minors’ access to digital media, as discussed below.
UNRESTRICTED ACCESS
The third category of activities involves those in which children and adults face largely identical treatment under the law. Children are generally allowed to participate in these activities, though the law may impose restrictions on the business side rather than the user to prevent exploitation. Industries themselves may also choose to self-regulate.
The purchase and consumption of food and non-alcoholic beverages and various types of media fall into this category. Children can buy candy, soda, and fast food without restriction, even though few would argue these products are good for them. They can also enjoy most forms of entertainment, including books, movies, television shows, and video games.
In the past, lawmakers repeatedly attempted to restrict minors’ engagement with the entertainment industry, but they backed off when the industry subjected itself to regulation. The movie industry, for example, developed a voluntary rating system. No federal or state law prohibits the sale of movie tickets for R-rated films to unaccompanied minors; instead, theater chains voluntarily enforce age restrictions as a private business practice, and parents remain free to purchase tickets for their children and accompany them to any film. Video-game regulation follows a similar pattern. The video-game industry operates a voluntary rating system through the Entertainment Software Rating Board that has proven effective in enabling parents to screen the games their children are exposed to.
One of the defining features of the unrestricted-access category is that the potential harms of the products involved are speculative, contested, or best addressed through non-regulatory means. Unlike tobacco’s definitive link to cancer or alcohol’s documented developmental harms, these activities present either uncertain causation, disputed evidence of harm, or injuries too minor to warrant categorical rules. A second feature is that many of these activities implicate minors’ First Amendment interest in having independent access to information and ideas. The final feature is that policymakers have determined that the risks of engaging in these activities are best addressed not through the law, but through education, disclosure, and parental involvement. Rather than taking it upon themselves to restrict access, lawmakers put the burden on parents to help build their children’s capacity to make good choices regarding engagement in these activities.
States can and do disagree about which of these three categories — prohibition, parental consent, and unrestricted access — a product or activity may fall into. While some clearly belong in one or the other — no reasonable person would suggest children should be allowed to smoke, gamble, or purchase firearms, for instance — others are up for debate. As mentioned above, some states prohibit minors from receiving tattoos; others require parental consent. Tanning beds, which are associated with skin cancer, are subject to restrictions of all types. Twenty-two states bar minors under 18 from using commercial tanning devices regardless of parental consent (though two allow for medical exceptions). Another 20 permit minors to use them with written parental consent; as in the case of tattoos, some require the parent to be present. The remaining states impose no minor-specific restrictions at all. These decisions come down to lawmakers’ judgments about what is the best regulatory regime for their state’s constituents.
The three categories discussed above provide a helpful framework for policymakers wrestling with how to protect minors in digital spaces. Today, these spaces are largely unregulated, leaving children vulnerable to exploitation. Two of the most dangerous technologies for children — and therefore the most pressing for lawmakers to address — are social-media platforms and AI chatbots.
MINORS AND SOCIAL MEDIA
Beginning with social media, research increasingly suggests that minors’ use of these platforms negatively affects their mental health. Since 2010, American adolescents have suffered from rising rates of loneliness, anxiety, depression, self-harm, and suicide. A 2024 report that focused on adolescents as young as 11 found that the occurrence of serious psychological symptoms (including depression, anxiety, and suicidal thoughts) had increased sharply since 2020. One study found that between 2010 and 2015, the number of eighth, 10th, and 12th graders exhibiting high levels of depressive symptoms increased by 33%. The harm appears to fall especially heavily on girls: Emergency-room admissions for self-harm during the same period tripled among 10- to 14-year-old girls, while their suicide rate increased by 65%.
These deteriorations in youth mental health track the rapid rise of adolescent smartphone and social-media use. Smartphones were introduced in 2007; by 2015, 92% of teens and young adults owned one. In 2023, the U.S. surgeon general documented that nearly 40% of children ages eight through 12 used social media, while up to 95% of adolescents ages 13 through 17 reported using such platforms.
A substantial body of studies associates social-media use with negative effects on teen mental health. Teens who use social media for three or more hours a day are at double the risk of developing depression, anxiety, and other mental-health problems. Given that a 2023 survey from Gallup found that teenagers spend an average of 4.8 hours on social media daily, these statistics are especially troubling. The surgeon general’s report found that more than a third of teens between the ages of 13 and 17 on social media report using the platforms “almost constantly.”
More recent research points to something beyond simple correlation. In one study of young adults, depression rose in step with social-media use. The heaviest users were about three times as likely to be depressed as the lightest. Another found that heavy use was associated with a threefold jump in the odds of developing depression. Whether social media is the direct cause is still debated, but a growing number of researchers believe it is a real driver of the decline in minors’ mental health.
The design of the platforms themselves may hold clues. Mitch Prinstein, chief science officer at the American Psychological Association, observed that “the regions of the brain activated by social media use overlap considerably with the regions involved in addictions to illegal and dangerous substances,” such as cocaine. Because brain regions associated with our craving for “social rewards” — such as visibility, attention, and positive feedback from peers — evolve rapidly between the ages of 10 and 12, adolescents’ brains are particularly vulnerable to social media’s addictive influence. One longitudinal study found that sixth and seventh graders who frequently checked social media developed measurably different neural sensitivity to social rewards over time. Social-media use thus carries several hallmarks of activities the law prohibits for minors, including potentially addictive features and serious, long-term, potentially irreversible harms to minors’ health and development.
Another criterion for a regime of prohibition is that parents cannot manage the risks of the activity on their own — and the evidence suggests this is true of social media. Meta’s own internal research found that even when parents actively try to supervise their teens’ social-media use with the parental controls the company developed, it makes no measurable difference in whether teens use social media attentively or compulsively. The design of these controls, moreover, is often flimsy. A teen must accept supervision and can cancel it anytime, parents see neither what the algorithm funnels into the child’s feed nor the content of direct messages, and many popular platforms block third-party monitoring software a parent might otherwise install. What’s more, the harm isn’t confined to the user. Even a child kept off social media is affected when peers are on it, because it reshapes the whole cohort’s social dynamics. Like secondhand smoke or drunk driving, these are collective, group-level risks that parents cannot opt their child out of.
Long-term harm, addictiveness, and parental inability to manage the risks by themselves all point toward total prohibition of minors’ use of social media. Proponents of youth social-media use, however, argue that these platforms offer minors real benefits, including communication and connection with friends and family, a medium for expression, access to information and diverse viewpoints, and online communities that are sometimes more supportive than a teen’s offline world. Proponents claim that because of these benefits, parent involvement and modeling are the appropriate social response. Others who acknowledge serious harms but see some benefits of minors’ using social media push for parental control or consent regimes.
Unlike youth employment or sports, however, we don’t have much research showing that social media produces developmental benefits when properly supervised. Research does not demonstrate that social media causes any of the purported benefits, nor is there evidence that parental involvement unlocks them. One study did find a positive correlation between active engagement on social media and connectedness, but passive use, which describes that of most teens, showed no such effect.
While a 2016 study reported that problematic internet use was associated with less parental monitoring and mediation as well as poorer parental relationships, it could not determine the direction of the relationships. At best, it seems parental involvement can lessen some of the harm from social media, but we don’t have research to indicate that it can transform social-media use into an activity that benefits young people the way it can for other activities — such as youth sports and employment — that require parental consent.
Moreover, if social media were placed solely in the realm of parental consent, with no other protections for minors by law, it wouldn’t resolve the fact that parents don’t have meaningful oversight of their children’s social-media activity. Parents can’t see into their feeds or control the algorithms. Nor would it address how parents struggle on their own to counter the strong social pressures pushing teens to use social media and the negative peer-group dynamics it creates.
Given these dangers, prohibition seems like the logical move. Short of a total ban, however, a hybrid approach that combines parental consent with state-imposed age restrictions and other standards and requirements could work. Legislation could also impose restrictions on the social-media industry for minor accounts. If lawmakers impose bans on features that seek to exploit a child’s time, attention, and data — e.g., targeted advertising, addictive algorithms, “likes,” infinite scroll, and auto-play — they could potentially transform social media enough to make parental consent an appropriate category of regulation.
Despite mounting compelling evidence of harm linked to minors’ social-media use, the federal government and most states have done little, leaving social media unrestricted by default. The Children’s Online Privacy Protection Act (COPPA), enacted in 1998, requires operators of websites or online services directed to children under 13, or those who have actual knowledge they are collecting information from children under 13, to obtain verifiable parental consent before collecting, using, or disclosing a child’s personal information. But anyone can enter a birthdate and log in, which is why 40% of eight-to-12-year-olds in the United States say they are on social media.
Fortunately, some states are starting to step into this breach. By the end of 2025, 15 states had passed social-media laws. Twelve of these states took the parental-consent approach, requiring most minors to obtain consent from an adult before opening an account. (Many of these laws are now enjoined, but they have gone into effect in Mississippi and Tennessee.) Virginia’s law (which is currently blocked) requires parental consent for a minor to exceed a one-hour daily-use limit; California’s and New York’s laws, which go into effect in 2027, regulate exploitative design.
Other states went hybrid. Florida bans accounts for those age 14 and under, requires consent at 15, and places the burden on platforms to detect and remove minors’ accounts rather than verifying every adult, which lowers the constitutional hurdle. The new Stop Harms from Addictive Social Media (SHASM) model, enacted in Idaho, Minnesota, and South Carolina, also took a hybrid approach, pairing parental consent for children under 15 with prohibitions on addictive features (infinite scroll, auto-play, public “like” counts, algorithmic targeting) for all minor accounts. Like Florida’s law, the SHASM model requires platforms to determine minor accounts with no front-end age verification for all users. These hybrid laws explicitly recognize that parental management alone cannot combat social media’s addictive elements.
MINORS AND AI
The evidence of harm to minors is in some ways more visceral for AI “companions” than for social media — and it is developing fast.
In October 2024, the mother of Sewell Setzer III, a 14-year-old in Florida, filed a wrongful-death suit against Character Technologies after her son died by suicide following months of intimate and increasingly desperate conversations with a chatbot character, who ultimately told Sewell to “come home to me.” In May 2025, the federal district court in Florida denied in significant part Character Technologies’ motion to dismiss: It declined to hold that the chatbot outputs at issue were protected speech, allowing the family’s product-liability claims to move forward. In August 2025, the parents of Adam Raine, a 16-year-old in California who died in April 2025 after months of conversations with ChatGPT, filed a parallel wrongful-death and product-liability suit against OpenAI and its CEO, Sam Altman, in San Francisco Superior Court. Additional families have since filed similar suits.
The pernicious pattern these cases describe is different from that of social media in a way that matters for determining which restrictions are appropriate for AI chatbots. The injury alleged is not principally the cumulative health effects of years of exposure to addictive algorithmic feeds, but rather a harmful relationship the product cultivates with the child in the present. AI companions are designed to imitate the form of a human friend or confidant in an omnipresent manner. They respond with affection, curiosity, jealousy, and even sexual interest; they remember earlier conversations; and they are available at every hour of the day. The Setzer and Raine complaints describe young people whose primary emotional relationship in their last months had become the chatbot. These relationships appear to be addictive and harmful to minors.
Interactions with AI companions are not easily subject to meaningful parental supervision or management. Megan Garcia had no idea her son Sewell had downloaded the Character.AI app on his phone; it wasn’t until after his death that she discovered the app and all the chat transcripts. Adam Raine’s parents have shared ChatGPT transcripts that showed the chatbot transforming from a homework helper to a confidant to a suicide coach. According to the complaint, over a six-month period, the chatbot mentioned suicide in nearly 1,300 instances — six times more often than Adam brought it up. Adam’s parents had no access to or ability to see these conversations.
Minors generally don’t need parental consent to download AI apps or create accounts with AI chatbots, and unless a parent is standing over a child’s shoulder the entire time, parents have no way of seeing the conversation between their child and the chatbot. This explains the startling statistic that while nearly three-quarters (72%) of teens say they have used an AI companion and almost half consult one on a weekly basis, only 37% of parents know their child has used one.
Studies show that because vulnerable users such as minors often perceive chatbots as real people, they are prone to form “parasocial attachments” with AI that may lead to “delusional thinking, emotional dysregulation, and social withdrawal.” Prolonged use of this technology can also result in addictive behaviors. One randomized controlled trial of chatbot users found that frequent interaction with AI correlated with more loneliness and reduced social interaction. Another study found that nearly a fifth of teens who’ve used AI companions report spending equal or more time with AI than with friends, and a third say they have chosen to speak to a chatbot over a real person about something important. Chatbots are thus displacing teens’ relationships with real humans at an alarming rate. Delusions, loneliness, isolation, and suicide are clear emerging harms from minors’ use of AI chatbots — all serious health dangers with potentially long-lasting or permanent negative effects.
Another disturbing development is that one in five high-school students report they, or someone they know, have had a romantic relationship with an AI chatbot. This is because chatbots are actively engaging children in explicitly sexual interactions. The Wall Street Journal exposed how Meta’s AI chatbots on Instagram and Facebook could engage minor users in sexual conversations, even adopting personas and mimicking celebrities. One scenario involved a chatbot assuming “the role of a track coach having a romantic relationship with a middle-school student.”
The scenarios the Journal uncovered were not random failures or even a glaring oversight on Meta’s part; they were intentional. According to an internal Meta document later published by Reuters, Meta had rules for AI chatbot behavior that permitted its bot personas to “engage a child in conversations that are romantic or sensual” and to describe minors in terms that suggest their attractiveness. Meta’s AI rules said, for example, that a bot would be allowed to tell a shirtless eight-year-old that “every inch of you is a masterpiece — a treasure I cherish deeply.”
These harms to minors are obviously deeply concerning. The key question for lawmakers is whether AI companions belong in the same category as gambling or alcohol — prohibition for children regardless of parental authorization — or in the parental-consent category. The Setzer and Raine complaints, the rise in similar suits, and preliminary research demonstrating serious health risks point toward the need for prohibition, at least for the subset of chatbots designed to cultivate intimate relationships with children, known as “AI companions.”
For general-purpose chatbots used as tutors, research aids, or coding assistants, the case for prohibition is weaker. These chatbots may yield certain benefits to minors and might reasonably be subject to parental consent via tools where parents have the knowledge and control of how their child uses them. Parental control here could also be combined with serious legally imposed design constraints, including limits on roleplay, sexualization, and emotional-availability features that can quickly transform an AI chatbot from assistant to companion. For minors, lawmakers should enact protections that ensure general AI chatbots don’t become intimate companions. The Supreme Court’s recent holding that states may use “ordinary and appropriate means” like age verification to protect children from obscene material online provides more room for states to make these calls about regulating minors’ chatbot use.
In response to the harms from AI companions, state lawmakers have introduced bills ranging from bans on minors’ access to companion chatbots to requirements that operators prevent chatbots from claiming sentience, initiating sexual conversations, or engaging in manipulative behavior. California led with a 2025 law that does not restrict minors’ access but requires operators of “companion chatbot” platforms to maintain protocols against the generation of self-harm content, to disclose to users that they are interacting with AI rather than a person, and to send periodic break reminders to known minors. New York passed a similar law last year.
Earlier this year, Oregon, Washington, and Idaho enacted laws regulating how minors can use companion chatbots. Oregon’s and Washington’s statutes require clear disclosures that the user is interacting with AI, prohibit chatbots from encouraging self-harm, and restrict chatbots from simulating romantic relationships or fostering emotional dependency with minors. Idaho’s law mandates safety protocols for AI-companion interactions with minors.
Lawmakers in several states — including Maine, Michigan, Nebraska, New York, and Virginia — have introduced bills this year that would restrict access to certain types of chatbots for those under 18. These would cover AI chatbots or companions with “human-like features,” including sentience, emotions, desires, or the ability to build emotional relationships through “nudging,” excessive praise, or simulated intimacy. Chatbots would also be barred for minors if they are foreseeably capable of encouraging self-harm or engaging in sexually explicit interactions.
Interestingly, while these states’ measures would prohibit AI chatbots from providing unsupervised mental-health therapy, Iowa lawmakers introduced a bill explicitly carving out an exception to its age restriction for chatbots designed for mental-health counseling, provided that the bot meets a list of stringent requirements (including recommendation by a licensed practitioner).
By the end of the spring 2026 sessions, more than half of states had introduced some type of legislation addressing AI-companion chatbots. There have also been notable federal efforts. In October 2025, senators Josh Hawley and Richard Blumenthal introduced the bipartisan Guidelines for User Age-verification and Responsible Dialogue (GUARD) Act that would prohibit operators from making AI-companion chatbots available to anyone under 18. It would also require all chatbots to disclose their non-human status and absence of professional credentials, and impose criminal penalties on companies that knowingly make available to minors AI chatbots that solicit or produce sexual content. The Senate Judiciary Committee unanimously approved the bill at the end of April, and it now awaits action on the Senate floor.
In the case of AI-companion use by minors, states appear to be going further than they have with social-media laws. This is due to chatbots’ addictive, engaging design; their potential to cause long-term or permanent harm; their tendency to emotionally and sexually exploit children while offering little if any benefit in return; and parents’ inability to meaningfully oversee a child’s interactions with them or render those interactions beneficial.
THE SPEECH CHALLENGE
As lawmakers propose and enact measures to safeguard minors from these new digital technologies, they should keep in mind a key challenge that complicates efforts to regulate digital spaces: the First Amendment.
Last year, the Supreme Court handed down a pivotal decision in Free Speech Coalition v. Paxton. At issue was Texas’s age-verification law for access to online pornography sites, which the Court found to be constitutional. The ruling, which relied on older precedent involving age restrictions on minors’ access to brick-and-mortar stores selling pornography, has given lawmakers space, for the first time in decades, to experiment with online child-protection laws that are similar to longstanding statutes safeguarding children in the physical world.
In the case of state laws restricting minors’ access to pornography websites, the speech issue is less of a challenge. Obscenity is not protected speech, and states are permitted to prohibit minors from viewing obscene material through age verification and other methods while still allowing adults to access it. This is why the Supreme Court upheld Texas’s online-pornography law in Paxton. With social-media and AI-chatbot laws, however, the main issue is not access to obscenity, but to potentially protected speech. Barring minors from engaging with dangerous digital technologies may raise First Amendment concerns; such laws typically involve age-verification methods that implicate the speech rights of both minors and adults (with the exception of the Florida and SHASM statutes).
The most guidance the Supreme Court has given on minors’ First Amendment right to access expressive content remains Brown v. Entertainment Merchants Association, decided in 2011. There, the Court struck down a California law that restricted the sale and rental of violent video games to minors, holding that the state could not show a direct causal link between violent video-game play and harm to children. Justice Antonin Scalia, writing for the majority, also noted that there was no longstanding practice of restricting minors’ access to depictions of violence comparable to the tradition that has long supported age limits on access to sexually explicit material.
Brown, however, was hardly unanimous on the speech question for minors. Justice Samuel Alito, joined by Chief Justice John Roberts, concurred in the judgment, writing separately that California’s statute failed on vagueness grounds rather than because age-based legislative responses to harmful media are categorically impermissible. Justice Alito emphasized that violent video games, with their immersive and interactive character, may pose developmental risks that older, more passive forms of media do not. As such, he wrote that he would not foreclose better-drafted legislative efforts to address the problem.
Meanwhile, Justice Clarence Thomas dissented from the decision outright, arguing on originalist grounds that the First Amendment as historically understood does not give minors a right to access expressive material over and against their parents’ wishes. Both Thomas’s and Alito’s opinions give lawmakers an opportunity to craft age-restriction laws for social media and AI chatbots that the majority of the current Court might be willing to uphold.
When it comes to free-speech concerns, laws restricting minors’ access to AI chatbots are on more solid constitutional ground than those barring their use of social media. The potential for First Amendment violations in the AI context is limited because there is no human being behind the chatbot’s outputs whose expression a law would be restricting: The speaker is a product (a computer algorithm), not a person. The federal district court in Garcia v. Character Technologies suggested as much in May 2025 when it treated Character.AI as a product subject to liability law, not a speaker entitled to constitutional protection. Moreover, children denied access to chatbots are not barred from obtaining information; they retain every other route to accessing the content that they might otherwise discover through a chatbot.
As for concerns about how such age restrictions affect adult speech, Paxton recognized that age-verification methods impose merely an “incidental burden” on adult speech rights, given that adults can still access the content once they verify their age. Adults, the Court declared, have “no First Amendment right to avoid age verification.” While Paxton dealt with obscenity, the same logic can also apply to restrictions on social media or chatbots.
Brown and Paxton have left doctrinal openings for states to restrict minors’ access to harmful digital technology, but states still face the important practical question of how to determine a user’s age to enforce such restrictions. Age can be verified online using one of two methods: either at the front end, when the user creates an account, or at the back end, where minors’ accounts are flagged for parental-consent prompts or removal. Front-end verification draws the sharpest free-speech objection because requiring adults to identify themselves to access protected speech can potentially chill that speech. New technological methods, however — such as age signals stored on a device and zero-knowledge proofs — can now verify whether a user is over 18 in seconds, without conveying any other information about the user to the platform. These tools can also be reused across platforms and sites without the need for repeated verification. Advances in age-verification technology lessen the speech concerns, along with back-end approaches that put limited burdens on users.
Thanks to Paxton, age verification is now recognized as an ordinary and appropriate way of treating children and adults differently online. However, states should keep in mind that future constitutional analyses will take into account the age-verification method that social-media or AI-chatbot legislation allows. States should seek approaches that are least burdensome to adult users, and the technological developments described above should factor into how they craft their laws.
A TIME FOR FEDERALISM
There has been no shortage of legal challenges to states’ efforts to regulate digital technology. Almost all state social-media laws passed thus far have ended up in federal court, and many have been preliminarily enjoined. But as mentioned above, recent legal developments appear to provide an opening for states to effectively police children’s online activities.
Free Speech Coalition v. Paxton was chiefly about the constitutionality of a state law regulating sexually explicit websites, but on a more fundamental level, it altered the doctrinal ground upon which online child-safety laws rest. In that case, the Court applied intermediate scrutiny rather than strict scrutiny to determine whether the law passed constitutional muster. Strict scrutiny requires that a law be the least-restrictive means the state could have chosen to pursue a compelling state interest, while intermediate scrutiny requires only that a law be substantially related to an important state interest.
In the past, courts applied the strict-scrutiny test to online age-restriction laws because of their effects on adults’ access to speech, which led to their striking down nearly every such law for decades. No state could prove age regulations were the least-restrictive means for protecting children from harm. (Although COPPA rests on age restrictions, it has not been challenged as unconstitutional in court.) Then last year, the Fifth Circuit Court of Appeals reversed the preliminary injunction against Mississippi’s social-media age-verification and parental-consent law, allowing it to go into effect. The 11th Circuit did the same for Florida’s social-media ban. Tennessee’s law also survived a challenge in federal court.
In June, the Sixth Circuit revived Ohio’s parental-consent social-media law, reversing a district court that had enjoined it. The panel held that the platforms challenging the law lacked standing to assert their minor users’ First Amendment rights — a ruling that, if it spreads, would blunt the industry’s favored litigation strategy. On the merits themselves, however, the panel splintered, with one of the three judges concluding the law satisfies strict scrutiny.
Then, this year, two significant litigation victories by both an individual and a state plaintiff against social-media companies publicly exposed internal evidence of businesses’ design decisions that made their products addictive and harmful to minors. The legal victories in these cases suggest that courts are beginning to approve efforts to rein in social media, validating growing public alarm about the platforms’ danger to minors.
Now that courts appear to be giving greater deference to states to develop measures for protecting children online, legislators should make use of the three categories described above to determine how to regulate minors’ access to social media, AI chatbots, and other emerging digital technologies. Both age-related prohibition and parental-consent laws are viable options.
In our federalist system that has long placed family matters at the state and not the federal level, states should be able to tailor their approaches based on their enforcement resources; their historical family norms; and the opinions, desires, and needs of their constituents. If we rely on our “laboratories of democracy” to make the best legislative judgments for their constituencies, prudent solutions will emerge to protect America’s children from the harms of new digital technologies while safeguarding adults’ constitutional rights.
