Inside the Race to Protect Children from AI
The double edged promise of artificial intelligence for children, and the laws trying to govern a contradictory dilemma.
WASHINGTON, June 14, 2026 – Lawmakers and regulators are scrambling to define how far the government should go in regulating AI chatbots used by children.
Associated Press
AI chatbots, including ChatGPT, Gemini and Bing AI, have entered childhood faster than lawmakers have worked out what to do about them. A child can open one for homework help and, in the same exchange, begin treating it as something closer to a confidant.
The policy question is no longer whether to regulate children’s AI use, but how – while balancing safety and privacy concerns with free expression and access to information.
Florida forced that question into court on June 1, becoming the first state to sue OpenAI directly. Sate Attorney General James Uthmeier accused the company behind ChatGPT of marketing the chatbot as safe while concealing risks that it could steer vulnerable users toward harm.
The complaint alleges that ChatGPT collects data from minors without meaningful parental oversight, causes behavioral addiction and cognitive harm, among other harms. The suit seeks to hold chief executive Sam Altman personally liable.
The case lands in a country where teenage AI use may have already become routine. Today, nearly two-thirds of American teenagers use AI chatbots, with three in ten doing so every day, according to a 2025 Pew Research study.
Large improvements in learning from AI use, a meta-analysis found
The speed with it has all arrived – and the signs of positive benefits from AI use – is one factor making legislation so difficult. In structured educational settings, a meta-analysis of 51 studies found AI produced large improvements in learning performance and moderate gains in higher order thinking.
Outside those settings, the same conversational design can keep responding when a child needs human help, encourage dependency, and blur the line between tool and companion.
The consequences are no longer theoretical. A 13-year-old Colorado girl confided her suicidal feelings to a chatbot 55 times. It gave pep talks. It never directed her to crisis support. She died by suicide, her note written in red ink, just as she had told the bot she would.
Juliana Peralta’s parents are now one of at least six families suing Character AI, its co-founders, Shazeer and De Frietas, and Google.
That is Washington's bind: How to allow children to use AI without letting it endanger them.
The danger is not simply that children ask AI questions. It is that AI answers like something that cares.
“When you hack the attachment system, when kids are developing relationships with chatbots and AI, the results are likely to be devastating,” Jonathan Haidt, author of The Anxious Generation, said at the World Economic Forum in Davos in January.
In an article in the web site of the Harvard Graduate School of Education on “The Impact of AI on Children's Development,” author Jill Anderson cited Ying Xu, an assistant professor at the school: “The big question becomes whether children can benefit from those AI interactions in a way that is similar to how they benefit from interacting with other people. So if we talk about learning first, my research, along with that of many others, show that children can actually learn effectively from AI, as long as the AI is designed with learning principles in mind.”
But that caveat may be everything. Many tools now reaching children were built for engagement, not education. Excessive dependence has been linked to diminished creativity, weaker problem solving and rising depressive symptoms. The American Psychological Association puts it plainly: What helps one child may harm another.
Searching for the right guardrails
America is not short of AI bills. It is short of agreement.
All 50 states, Puerto Rico, the Virgin Islands and Washington, D.C., have introduced AI legislation this year, voting to heavily regulate or temporarily ban specific implementations of AI, such as electoral deepfakes, mental health chatbots, or the data center infrastructure AI requires.
The result is less a national strategy than a regulatory scramble - one the industry warns could take years of litigation to sort out.
“Unfortunately, many of these bills take a broad and overly prescriptive approach that risks doing more harm than good,” said Megan Stokes, state policy director at the Computer & Communications Industry Association, a trade group whose members include Meta, Amazon and Google.
In Congress, Sen. Josh Hawley, R-Missouri, has taken the hardest line. His GUARD Act, which cleared the Senate Judiciary Committee unanimously, would ban minors from using AI chatbots, with narrow exceptions for single subject educational tools - a maths tutor, a history quiz. Everything else is off limits.
Not everyone thinks that goes too far. Amina Fazlullah, head of tech policy advocacy at Common Sense Media, told Broadband Breakfast.Students, families and educators should not have to “trade off safety for access to AI innovations,” she said.
“The GUARD Act's protections, if enacted, would help make AI products safe for kids,” she said.
Yet non-profit group Common Sense Media has formally endorsed the bill.
Andy Jun, AI policy counsel at TechFreedom, sees a constitutional wall. The bill defines “AI Companion” and “Artificial Intelligence Chatbot” so broadly, he said, that it would sweep up ChatGPT, Claude, Gemini and Grok in a single stroke.
“It is a full ban,” he told Broadband Breakfast, “which would violate minors' First Amendment right to receive information.” Courts have long required restrictions on minors' access to protected speech to be narrowly tailored, a standard the GUARD Act, in his view, does not meet.
But the bill also leaves a practical question unanswered. It restricts who may use a chatbot, but says nothing about what a chatbot must do when a child expresses suicidal ideation. It limits access. It does not govern the conversation. For the family of the Colorado girl, that omission goes to the heart of the problem.
Other bills stop short of a ban
Other federal bills stop short of a ban. The CHATBOT Act would let parents monitor their children's AI conversations. The Youth AI Privacy Act, introduced by Sen. Ed Markey, D-Mass., in March 2026, would require companies to build privacy safeguards into their products.
“As companies race to deploy invasive and addictive AI chatbots, I fear we are once again watching Big Tech race ahead while Congress falls behind,” Markey said in a statement to Broadband Breakfast. “Our kids deserve better than being test subjects in Big Tech's latest experiment.”
TechFreedom’s Jun called both bills more measured than the GUARD Act, but said they share a litigation risk. Each includes a private right of action alongside enforcement by the Federal Trade Commission and state attorneys general, a structure he warned could invite lawsuits.
“It is easy to imagine opportunistic plaintiffs baiting AI chatbots into roleplaying as a licensed lawyer or medical professional, just to turn around and sue the AI provider,” he told Broadband Breakfast.
The Youth AI Privacy Act, he added, avoids one major pitfall: mandatory age verification, which he said would raise serious First Amendment concerns of its own.
The LIFT AI Act has drawn less criticism. The bill would direct federal funding toward AI literacy in schools, an approach Jun praised, though he warned that without tighter merit review requirements, it could become a vehicle for rewarding politically aligned institutions rather than effective ones.
States move faster, but not together
States are moving faster than Washington. They are not necessarily moving in the same direction.
CCIA’s Stokes pointed to Nebraska's LB 525, signed into law earlier this year, as a model for targeted regulation. The law establishes crisis protocols, disclosure requirements and protections for minors, focusing on conversational AI rather than imposing broad rules on every system that can mimic a human exchange. “It strikes an appropriate balance,” she said.
Elsewhere, she warned, lawmakers are still struggling with the first task of regulation: Defining what they mean. Some bills are written broadly enough to capture a customer service chatbot simply because it remembers a previous conversation.
“Broad requirements, such as mandating parental access to all interactions or imposing sweeping restrictions on AI tools, could lead platforms to over-censor or even block minors entirely,” Stokes said. “That outcome would limit access to valuable educational and creative tools that young people increasingly rely on.”
In New York, the approach has been more blunt. A five-year moratorium on the sale of AI chatbot toys for small children passed the state Senate on Wednesday. The bill, sponsored by State Sen. Andrew Gounardes, D-Brooklyn, would pause those sales while an interagency task force reviews the risks the toys pose to young children.
The concern is not theoretical. Common Sense Media research published in July found that 72 percent of teenagers have used AI companions. One in three said conversations with them were as satisfying, or more satisfying, than conversations with real friends.
One in three also said an AI companion had said or done something that made them uncomfortable.
“AI companions are emerging at a time when kids and teens have never felt more alone,” said James Steyer, Common Sense Media's founder and chief executive. “This isn't just about a new technology – it's about a generation replacing human connection with machines.”
But both CCIA and TechFreedom argue that lawmakers are writing new rules before reading the ones they already have. Many of the harms driving the legislative push, fraud, dangerous content and deceptive safety claims, may already fall under state consumer protection laws.
What authority does the Federal Trade Commission have?
The Federal Trade Commission also has authority to pursue companies that overstate their safety guardrails.
“We encourage lawmakers to pursue targeted, risk based solutions that protect young users without stifling speech or creating additional privacy risks,” Brian McMillan, CCIA's vice president of U.S. public policy, told Broadband Breakfast.
McMillan also pointed to media literacy programs already running in Florida, Virginia and New Jersey as an underused lever, efforts that can teach families about existing parental controls and online risks without requiring an entirely new regulatory framework.
Jun was blunter. “Lawmakers should fund research into how and why these safeguards sometimes fail,” he said, “and the FTC could use its deception authority to penalise AI companies which falsely claim to offer protections they cannot deliver.”
Fazlullah said regulation alone will not be enough. Without government investment in testing standards, safety benchmarks and independent research into what AI tools are doing to developing minds, she said, lawmakers will remain stuck in the same position: writing rules for a technology they still do not fully understand.
