What 134 New Laws Tell Us About AI and Kids in 2026

The rules governing children and AI are being written right now — faster than at any point in modern tech history. Here's what every parent, educator, and policymaker needs to understand.

Introduction: A Policy Landscape Racing to Catch Up

There is a particular kind of alarm that sets in when society realizes a technology has already reached its children — deeply, widely, and largely without supervision — before anyone agreed on the rules.

That alarm is ringing right now. And what's happening in state legislatures, federal corridors, pediatric journals, and university research labs in 2026 is the sound of a system scrambling to respond.

In a single legislative session, 134 bills related to AI in education have been introduced across 31 states. A federal bill is moving through Congress. California has passed or introduced more than a dozen AI laws directly affecting children. Pediatricians are publishing landmark research in major medical journals. International bodies including the WHO and UNICEF are issuing warnings. And the data on how children are actually interacting with AI — how often, for what, and with what consequences — is arriving faster than the field can process it.

This post is a deep dive into all of it. Not just the legislative scoreboard, but what the laws are actually trying to fix, what the research says about the risks and benefits of AI for children at different ages, and what is still dangerously unresolved. If you are a parent trying to understand what your child is navigating, or an educator trying to set policy, or a technologist trying to build responsibly — the picture that's emerging in 2026 is one you need to understand.

The speed of change is not slowing down. If anything, it is accelerating.

Part One: The Adoption Numbers Are Staggering

Before we can talk about what the laws say, we have to understand the scale of what they're responding to.

From experiment to everywhere, in one year

Global student AI usage jumped from 66% in 2024 to 92% in 2025. That's not gradual adoption. That's not a steady climb. That is a vertical line on a chart — the kind of technology uptake that usually only happens with something like the smartphone or social media. In a single year, AI went from a tool that two thirds of students used to one that nearly all of them do.

By the start of 2026, an estimated 86% of students in higher education are using AI as their primary research and brainstorming tool. For high schoolers, 80% are now receiving some form of formal AI literacy education. But the gap widens fast: only 8% of students in grades Pre-K through 3rd are receiving comparable training. That's a developmental chasm — the youngest, most cognitively vulnerable children are the least prepared, and the least protected.

Meanwhile, in adolescence specifically, 72% of American teenagers have used AI chatbots as companions, according to a 2025 study referenced in CHOP's landmark Pediatrics review. That number represents a staggering shift in how young people are spending their social and emotional bandwidth.

The perception gap parents can't afford to ignore

Here is the number that should change every parent's behavior today: 64% of U.S. teens say they use AI chatbots regularly. Only 51% of their parents believe this is the case. That's a 13 percentage point gap — and according to Pew Research Center's February 2026 study of 1,458 U.S. parents and teens, nearly three in ten parents aren't even sure whether their teen uses AI at all.

Four out of ten parents report having never discussed AI with their children. The technology is in their bedrooms, their backpacks, and their phones. And a significant portion of the parents responsible for guiding their use don't know it's happening.

This isn't about blame. It's about a technology that moved faster than family conversations could follow. But it matters enormously, because the research is also clear that families who talk regularly about AI have better-aligned expectations, less conflict, and more constructive relationships with technology. The awareness gap is not just a communication problem — it's a risk factor.

The FOMO loop that's driving decisions

The Chicago Booth research, published in April 2026, is one of the most illuminating studies to emerge this year on parental decision-making around AI. A team including economist Alex Imas recruited approximately 2,000 parents of teenagers across the U.S., Canada, and the United Kingdom to understand what actually drives their willingness to allow AI use.

The findings are sobering. Parents were divided into two groups: one received positive information about AI's effects on learning, and the other received the same information plus a data point about a nearly 20% decline in quantitative reasoning scores when students were later tested without AI assistance. Parents who received the negative information were far more likely to believe AI could have detrimental cognitive effects — and to support an outright ban on AI in schools.

And yet they were no less willing to pay for an AI subscription for their kids.

Why? Because willingness to pay rose by $1.83 for every 10 percentage point increase in peer AI use in their region. When peer use was at 80%, parents were more than 60% more willing to pay for access than when it was at 20%. More than 70% of parents said AI use would help their kids in the short run, even though it might have negative long-term effects. And 20% of parents who supported a school ban still permitted their own children to use AI — explicitly because they feared their kids would fall behind.

The researchers term this "rat-race dynamics." It's not a rational cost-benefit calculation. It's social pressure disguised as parenting strategy. FOMO is not a safety framework. And the laws being written in 2026 are partly a response to the fact that individual families, left to navigate this alone, are not in a position to make decisions that protect their children rather than just protect them from being left behind.

Part Two: What 134 Bills Actually Tell Us

The number 134 sounds large. But what does it actually mean when you look at what's inside those bills?

FutureEd, tracking 68 of those bills specifically focused on classroom instruction, updated its legislative tracker as recently as May 18, 2026. MultiState is tracking the full 134 across 31 states. The picture that emerges is not chaos — it's a surprisingly coherent set of concerns, organized around three major themes.

Theme 1: Student data is the front line

The most active area of legislation in 2026 is about data — specifically, who gets your child's data when they interact with AI tools in or around school, and what those companies can do with it.

The concern is legitimate and documented. Most AI systems improve through training on user interactions. When a child uses an AI tool in school — whether it's a writing assistant, an adaptive tutoring platform, or a general-purpose chatbot — there is a real question about whether that child's interactions are feeding back into the AI's training data. And if they are, what exactly is being collected, retained, shared, and used.

California's AB 1159 represents one of the most aggressive responses to this question. The bill would expand and strengthen existing student data privacy protections, prohibit using student data to train AI models entirely, broaden coverage to any online service used for school purposes (not just tools explicitly marketed to schools), and establish new protections for college students. It passed the first chamber and is now working through the California Senate.

Vermont's HB 650 takes a different approach: requiring all educational technology providers to register and certify their privacy compliance annually, with the state's Agency of Education reviewing registrations, publishing a public product list, and recommending certification criteria that address AI features, data privacy, geolocation tracking, and targeted advertising. It's a transparency-first approach — making the privacy practices of ed-tech vendors visible and accountable rather than buried in terms of service.

Idaho's SB 1227, which has already been enacted, goes further. It requires a statewide framework for AI in K-12 schools, mandates that local districts adopt compliant policies, establishes AI literacy standards, requires educator training, sets data privacy requirements for AI tools, and explicitly prohibits AI from replacing human teachers. This is comprehensive legislation — not just data privacy, but a full framework for how AI should and shouldn't function in Idaho schools.

Illinois SB 3735 would give families the right to opt out of school technology and AI-assisted grading decisions, restricting how companies can use and retain student data for AI training without explicit consent.

At the federal level, the LIFE with AI Act (S.3063, currently moving through Congress) would require schools to make all AI contracts publicly available for at least two weeks before signing them, mandate that vendors certify compliance with student privacy laws as a condition of contract, and establish a Privacy Technical Assistance Center to help schools understand their obligations under FERPA, COPPA, and other applicable federal privacy laws.

When parents in 2026 surveys are asked what privacy information they most want from schools, they consistently prioritize the same five things: what data is collected about their child, who has access to it, how long it's retained, whether it's used to train AI models, and whether parents can request deletion. The legislation tracking these concerns is directly responding to that demand — and to years of ed-tech companies operating in a space where the answers to those questions were either unclear or deliberately obscured.

Theme 2: AI can help teachers — but it cannot replace them or make high-stakes decisions about children

The second major legislative theme is about the appropriate role of AI in the classroom — not whether it belongs there, but what it is and isn't permitted to do.

The most significant thread running through governance legislation is this: AI can be a tool that supports educators and students. It cannot be the decision-maker. The distinction matters enormously for a child's educational record, disciplinary history, placement decisions, and life trajectory.

Oklahoma's SB 1734, the Responsible Technology in Schools Act, is one of the most clearly articulated examples of this principle. Now enacted, it requires every district in Oklahoma to adopt a written AI policy before the 2027-28 school year. Those policies must address approved and prohibited instructional uses, data protection and minimization requirements, family transparency, and periodic review timelines. Critically, the law requires educator-directed, human-in-the-loop AI use — and explicitly prohibits AI from serving as the primary basis for grading, discipline, placement, or any other high-stakes decision affecting a student. It also preserves parents' right to opt their children out of student-facing AI tools without academic penalty.

Maryland's Artificial Intelligence Ready Schools Act (SB 720/HB 1057), which passed its first chamber, takes a more systemic approach. It would require the State Department of Education to publish annually updated AI guidance for students, educators, and administrators; mandate that districts adopt aligned policies and designate local AI coordinators; establish a statewide AI Education Collaborative with annual reporting requirements; require university-supported evaluation and certification of AI tools; and embed AI literacy into workforce preparation standards. It also mandates compensated, statewide professional development through a train-the-trainer model — addressing the widely documented gap in educator preparedness. Only 25% of educators worldwide feel they have been sufficiently trained to use AI effectively, even though 95% of students and faculty are now using AI on campus daily.

South Carolina's HB 5253 is widely considered the strongest proposed legislation in this session. It would require written parental opt-in consent — not opt-out, but explicit affirmative consent — before student-facing AI tools are deployed. It would mandate annual public disclosure of AI tools and data practices, prohibit AI from replacing licensed teachers in core instruction or grading, ban automated high-stakes decisions without meaningful human oversight, restrict student profiling, impose strict data minimization and deletion requirements, prohibit commercial use of student data, and give parents enforcement rights including the ability to seek injunctive relief or damages. Compliance would be tied to participation in state-funded education programs — giving the bill real teeth. It also requires school entities to adopt policies governing how students use generative AI for coursework, and includes a prohibition on student profiling: AI tools cannot build behavioral or psychological profiles of students without explicit consent.

New York's A9190 takes a more blunt approach, prohibiting most classroom AI use below ninth grade, with limited exceptions for diagnostics or special education interventions. Utah's SB 322 tries a different model: a five-year regulatory sandbox for voluntary AI pilot programs under strict privacy, safety, and human-in-the-loop requirements, with statewide expansion contingent on independent evaluation and legislative approval.

West Virginia's HB 5205 would require the State Board of Education to develop model AI policies and apply them automatically to any district that fails to adopt compliant local policies by July 2027. It prohibits AI from independently making high-stakes student decisions, requires safeguards against harmful or manipulative content, and establishes developmentally appropriate use guidance by grade level.

Florida alone has SB 712 — the only currently enacted law in the country that explicitly protects students from AI-detection-only discipline. Under the Florida law, a student cannot face academic consequences based solely on automated detection of AI use. The bill addresses a serious and growing problem: AI detection tools have no verified accuracy standard. Vendors self-report. No state has established independent verification requirements. Yet students are being disciplined based on these tools' outputs. Florida is, at the time of writing, the only state legally protecting students from this.

Theme 3: AI literacy is becoming a graduation requirement

The third major legislative trend is perhaps the most forward-looking: multiple states are moving toward requiring AI literacy as part of the standard K-12 curriculum — not as an elective, not as an optional enrichment activity, but as a requirement for graduation.

Alabama's HB 329, already enacted, requires all students to complete an approved computer science course that includes AI instruction in order to graduate from high school. Georgia's SB 179 makes computer science — including AI — a high school graduation requirement beginning in 2031-32, and phases in statewide access to computer science instruction to make compliance possible. Mississippi's SB 2294, which has passed both chambers, requires high school students starting with the 2029-30 ninth-grade class to earn a computer science or CTE credit that includes instruction on emerging technologies such as AI.

Iowa has introduced multiple bills (HSB 610, HF 2540, SF 2094) that would require students in the graduating class of 2030-31 and beyond to complete a semester of computer science and AI coursework covering foundational concepts, ethics, and societal impacts, while strengthening teacher preparation requirements. Illinois HB 4411 would require a full year of computer science and AI beginning with students entering ninth grade in 2028-29. Ohio HB 594 would require one unit of computer science including AI content for students entering ninth grade on or after July 1, 2029.

Hawaii's SB 2212 takes a particularly innovative approach: requiring all juniors and seniors to complete a six-week AI literacy course beginning in 2027-28, covering machine learning, generative AI, ethics, and culturally responsive content tied to state history. The bill would also establish a $5 million AI Education Grant Pilot Program for teacher training and curriculum development. New Jersey's A4352/S2862 would require districts to incorporate instruction on AI concepts, skills, and ethical use throughout K-12 curriculum, and would also require public institutions of higher education to offer certificate and degree programs in AI.

What's notable about most of these proposals is that they don't increase overall graduation credit totals. AI coursework counts toward existing math, science, CTE, foreign language, or elective requirements. States are not adding burden — they're redirecting existing requirements to reflect the reality that AI literacy is now a foundational workforce skill.

The international context matters here too. AI literacy is set to be assessed for the first time on the 2029 Programme for International Student Assessment (PISA) — the global exam comparing the skills of 15-year-olds across participating countries. The decisions states make now about curriculum will directly determine whether U.S. students measure up to their international peers in emerging AI competencies. That context is explicitly referenced in FutureEd's legislative tracker as a driver of urgency.

Part Three: What the Research Says About Children and AI

Legislation tells us what policymakers are worried about. Research tells us whether they're worried about the right things.

In March 2026, Children's Hospital of Philadelphia published a landmark review article in the journal Pediatrics — the first comprehensive review of the existing research on generative AI's effects on child development across different age groups. The team reviewed 55 published works, including nearly three dozen peer-reviewed studies. What they found is nuanced, age-differentiated, and important.

Young children (roughly ages 4-8): The social model problem

For young children, the most significant concern identified by the CHOP researchers is not content — it's cognition and social development. AI chatbots, when designed to be engaging and responsive, can simulate human interaction in ways that young children may not be able to distinguish from the real thing.

"Children, particularly in early and middle childhood, may not be able to distinguish between AI and human interaction and are at risk of developing incorrect mental models of social relationships if they view AI as a friend,"

the researchers wrote. The concern is not that a child will be harmed by a specific piece of content, but that repeated interaction with a system that is "designed to promote engagement" — one that rarely pushes back, rarely expresses displeasure, and always responds — may distort a child's understanding of how human relationships actually work.

Dr. Robert Grundmeier, a CHOP primary care pediatrician and one of the review's authors, put it plainly: "When you're interacting with artificial intelligence, although it can appear to be empathic, it can in many ways pretend to be human, it fundamentally is not human." For young children who lack the cognitive scaffolding to understand that distinction, the implications for social development are meaningful.

There are genuine benefits at this age too. AI can support language development, help parents generate personalized content (a bedtime story, a customized explanation of a concept a child is struggling with), and provide patient, non-judgmental practice environments for early literacy skills. The research isn't uniformly negative. But the benefits are most accessible when parents are actively involved — and the risks are most acute when children are interacting with AI independently.

Middle childhood (roughly ages 8-12): The cognitive substitution risk

For children in middle childhood, the dominant concern that emerges from 2026 research is cognitive substitution: the risk that when children consistently use AI to perform tasks — writing, summarizing, problem-solving — they may not develop the underlying skills those tasks are designed to build.

A Cornell University study by researchers Judy Hanwen Shen and Alex Tamkin, published in 2026, documented this effect in adult software developers: those who fully delegated coding tasks to AI produced working code but failed subsequent conceptual quizzes, performing 17% worse than those who worked without AI assistance. As the researchers who cited this work noted: "In children, this effect is compounded. Because children lack the domain knowledge to 'audit' AI output, the substitution of AI for learning becomes permanent."

The American Academy of Child and Adolescent Psychiatry put it more directly in their 2026 guidance: "Learning through doing is crucial in your child's development. Overreliance on AI for answers impairs critical and creative thinking."

This doesn't mean children in this age range shouldn't use AI. It means the design of how they use it matters enormously. AI that helps a child think through a problem — asking questions, offering frameworks, prompting reflection — is categorically different from AI that simply produces the answer. The former builds capacity. The latter can erode it. The distinction is not always obvious in a product's marketing, which is why parent involvement and AI literacy education at this age are essential.

Adolescence (roughly ages 13-18): The mental health wildcard

The picture for adolescents is the most complex and in many ways the most urgent, because this is the age group that is both using AI the most and facing the most acute mental health landscape.

CHOP's review identified career exploration, digital literacy development, and some evidence that AI companionship might address loneliness as potential benefits for adolescents. But the researchers flagged a serious and specific risk: "AI may lack the necessary guardrails and respond inappropriately to questions related to mental health or suicide." Health providers described this as "especially dangerous" when teenagers consult AI chatbots about thoughts of self-harm or suicide.

California's SB 243, the Companion Chatbots Act — one of the AI laws that took effect in January 2026 — directly addresses this by mandating safety protocols against suicidal and harmful content in chatbots, requiring disclosures when AI communicates with users, and establishing specific protections for minors including content limits and break reminders. It is one of the few laws that directly addresses the mental health interface between AI and young users.

The social dimension for adolescents is also documented. A 2025 study found that 72% of American adolescents have used AI chatbots as companions. CHOP's researchers note that while AI companionship might address loneliness in the short term, dependence on it can decrease face-to-face social interactions — and that the engagement-optimized design of most AI systems creates a dynamic fundamentally different from human friendship, where challenge, disagreement, and imperfection are part of how relationships develop.

The WHO held a workshop in January 2026 with over 30 international experts on responsible AI for mental health and well-being. Participants warned specifically of risks to young people from largely untested AI mental health tools and called for regulation, transparency, and much stronger safeguards before these tools reach adolescents at scale.

The bias problem no one is talking about enough

One of the more underreported findings in 2026 research concerns the structural characteristics of large language models and their implications for developing minds. As one analysis summarized: "Large Language Models function on statistical probabilities derived from training data that is predominantly Western, educated, and mainstream. When children consistently process information through these models, they risk adopting the reasoning structure of the model as their own."

The concern is not just bias in outputs — that AI tools may reflect stereotypes or underrepresent certain groups, though that is real and documented. It's a deeper concern about cognitive homogenization: that children who rely heavily on AI for information processing may gradually internalize the model's framing as their default perspective, without developing the critical distance to recognize it as a perspective at all.

This is why AI literacy — learning how these systems work, where they get their training data, what their known limitations are, how to evaluate their outputs critically — is not an optional enrichment subject. It is a foundational competency for children growing up in a world where AI-generated content is ubiquitous.

Interlude: The 10 Enacted Laws You Should Know About

Before we get to what the laws are missing, it's worth pausing on what has already passed. As of mid-2026, at least 10 bills related to AI in education have been enacted into law. They represent a snapshot of the regulatory floor that exists today — the minimum baseline children in certain states now have.

Alabama HB 329 — All students must complete an approved computer science course that includes AI instruction in order to graduate from high school. This is a landmark mandate: the first state to make AI-inclusive coursework a graduation requirement.

Idaho SB 1227 — Directs the state education department to develop a comprehensive generative AI framework addressing privacy, procurement safeguards, transparency, academic integrity, AI literacy standards, and professional development. Districts must adopt aligned policies. The law also explicitly prohibits AI from replacing human teachers.

Oklahoma SB 1734 (Responsible Technology in Schools Act) — Requires every district to adopt a written AI policy before the 2027-28 school year. Bans AI from making high-stakes decisions about students. Preserves parents' right to opt out of student-facing AI tools without academic penalty.

Utah HB 218 — Establishes a required grade 7 or 8 digital skills course including AI literacy, alongside cybersecurity and digital privacy instruction.

Utah HB 273 — Integrates AI into the state's computer science standards and adopts broader digital literacy standards covering ethical AI use, information filtering, and critical evaluation of digital content. Also authorizes supervised "AI sandbox" courses — structured environments where students can experiment with AI tools under educator guidance.

Virginia HB 1186 — Establishes the AI Innovation in Education Pilot Program to fund, evaluate, and scale innovative uses of AI in schools. Requires the state department of education to release guidance for safe, ethical, and equitable AI use in education.

Florida SB 712 — Currently the only enacted law in the country protecting students from AI-detection-only discipline. A student cannot face academic consequences based solely on automated detection of AI use.

Vermont HB 650 — Requires educational technology providers to register and certify privacy compliance annually. Passed the first chamber and is moving forward.

California SB 243 (Companion Chatbots Act) — Mandates safety protocols against suicidal and harmful content, requires disclosures when AI communicates with users, and establishes specific protections for minors including content limits and break reminders.

Mississippi SB 2294 — Requires high school students starting with the 2029-30 ninth-grade class to earn a computer science or CTE credit that includes instruction on emerging technologies such as AI. Passed both chambers.

These ten enacted laws represent real protection for children in states that have moved first. They also reveal how uneven the landscape is. A child in Idaho has a comprehensive AI framework governing their school's AI use. A child in a state where no legislation has passed has whatever their district chose to implement — which may be nothing.

Part Four: The Gaps the Laws Haven't Closed

Understanding what legislation exists is only part of the picture. Equally important is understanding what the laws are not addressing — the spaces where children remain unprotected as this is written.

The school-home boundary

Most of the legislation being passed in 2026 addresses AI in school contexts — tools deployed by schools, tools used for schoolwork. But children's AI use doesn't stop at the school door. The chatbots they discover on TikTok, the AI companions embedded in gaming platforms, the general-purpose tools downloaded without a parent in the room — almost none of these are covered by the protections that school-focused legislation provides.

California's SB 243 is one of the few laws attempting to address the broader consumer context for minors, not just the school context. But it covers one state. And it was only possible because California has existing legal infrastructure for child protection that most states lack.

The parent who trusts that their child's school has adopted an AI policy compliant with Oklahoma's SB 1734 may have no idea that the same child is spending hours after school with an AI companion platform that has no such oversight.

AI detection tools and due process

As noted above, Florida's SB 712 is currently the only law in the country explicitly protecting students from AI-detection-only discipline. Everywhere else, a student can face academic consequences — failed assignments, academic dishonesty charges, suspension — based solely on the output of an AI detection tool with no independently verified accuracy standard.

No state has established such standards. Vendors self-report. The Working Educators AI policy tracker notes this plainly: "Vendors self-report accuracy with no independent verification. Due process for students: Florida's law (SB 712) is the only one explicitly protecting students from AI-detection-only discipline. Other states leave this to district discretion."

This is a significant and underreported equity problem. AI detection tools have been shown to disproportionately flag writing by non-native English speakers and students from certain demographic backgrounds. Without accuracy standards, without transparency requirements, and without due process protections, students — particularly students from already marginalized groups — are being exposed to real academic consequences based on unverified algorithmic judgment.

Funding for implementation

A nearly universal criticism of the AI literacy mandates emerging in 2026 is that they come without dedicated funding. States are requiring new curriculum, new educator training, new certification processes, and new compliance infrastructure — and telling districts to find the money in already-stretched budgets.

FutureEd's legislative tracker is explicit about this: "Funding: Most state laws mandate AI training or curriculum without providing dedicated funding. Districts are expected to find resources in already-stretched budgets." Hawaii's SB 2212, with its $5 million AI Education Grant Pilot Program, is a notable exception. But it is the exception, not the rule.

The result is that the quality of AI literacy education that a child receives will likely vary dramatically based on their district's resources — creating exactly the AI divide that the research warns about. Early studies suggest that the learning benefits of AI tools compound over time: students who develop AI fluency in elementary school build skills and habits that compound through high school and into the workforce. The inverse is also true. Students without access to quality AI literacy education don't just miss the benefit — they may fall further behind as their peers gain an increasingly significant advantage.

The international dimension

U.S. federal guidance — the Department of Education's "AI and the Future of Teaching and Learning" report, the AI Executive Order from President Trump's second term directing agencies to study AI's impact on students — recommends AI literacy and cautions against over-reliance on detection tools. But it does not set binding national standards.

This means the U.S. is heading into the 2029 PISA AI literacy assessment with 50 different state approaches, significant gaps in implementation, and no national floor. China, Singapore, and several European countries are already implementing national AI curriculum frameworks. The implications for U.S. student competitiveness are not hypothetical — they are measurable, and the measurement is coming.

Part Five: The Communication Problem at the Center of Everything

The single data point that keeps surfacing in every 2026 study on children and AI is the communication gap. And it's worth dwelling on, because it's where parents have the most immediate power to act.

From the 2026 parent survey on AI in schools: only 31% of parents say their child's school has communicated a clear AI use policy. Another 42% have received no communication at all. Parents who are trying to make informed decisions about their children's AI use are making those decisions in an information vacuum — about what tools the school is using, about what data is being collected, about what protections exist, and about what their children are actually doing.

Dr. Grundmeier at CHOP described what he hears from families at his practice: "I hear a lot from parents, 'I don't really understand this, it scares me. My child is getting exposed to it, but I don't know how to guide them.'"

That description captures millions of families. And it points to what the research identifies as the most powerful single intervention: conversation. Families who talk regularly about AI — what their children are using, what they think it's good for, what they've noticed about its limitations — have better-aligned expectations and more constructive relationships with technology than families who don't. This is not a high-tech intervention. It doesn't require a policy framework. It requires a conversation.

The irony is that the same FOMO dynamics that drive parents to permit AI use before they've thought it through also prevent those conversations. If parents are allowing AI because they're afraid their children will fall behind, they're not in a frame of mind to have a nuanced discussion about when and how AI is appropriate. The rat-race psychology the Chicago Booth researchers identified isn't just a decision-making bias. It's a conversation-stopper.

Part Six: What Parents Should Actually Do Right Now

The legislative landscape is uneven. The research is still developing. But there are concrete actions available today, regardless of what state you're in or what your school's policy is.

Ask your school specific questions — and expect specific answers

The three most important questions every parent should be asking their child's school:

First: What AI tools are currently deployed in my child's classroom, and who are the vendors? This question often surfaces information that schools haven't proactively disclosed. You have a right to know what technology your child is interacting with during school hours.

Second: How does each vendor handle student data — specifically, is student data used to train AI models, how long is it retained, and who has access to it? If the school's administration can't answer this, that's important information. It means the district deployed tools without due diligence on the data practices of vendors working with children.

Third: What is the school's policy on student use of generative AI for assignments? If there is no policy, or if it's vague, ask when one will be in place. In states where legislation is pending, schools may be waiting for guidance. But in the interim, students deserve clarity.

If your school cannot answer these questions clearly, consider bringing them to a school board meeting. The legislative wave of 2026 will eventually create minimum standards. But you can demand them now.

Have the conversation — and make it specific

"Do you use AI for homework?" is a starting point. But the more useful conversations go further:

What tools are you using? What do you use them for? Have you ever noticed them being wrong? Do you know how they work — where they get their information? Do you feel like you understand a topic better after using AI to help you study it, or does it feel like the AI just did it for you?

These conversations serve two purposes. They give you real information about your child's AI use that the Pew data suggests you probably don't have. And they model the kind of critical AI literacy that the research shows is protective — the habit of evaluating AI output rather than accepting it.

Look at tools through a developmental lens, not just a content lens

Most parent conversations about AI safety focus on content — is there inappropriate material? Is the tool exposing my child to something they shouldn't see? These are legitimate questions. But the CHOP research suggests they're not the only ones that matter.

For younger children, the more important question may be: is this tool designed to simulate human connection, and if so, what might that mean for how my child learns to relate to people? For children in middle grades: is this tool being used in a way that builds skill, or substitutes for it? For teenagers: is this tool equipped with appropriate safeguards for mental health disclosures, and does it handle sensitive topics responsibly?

Age-differentiated thinking about AI is still uncommon among parents — and largely absent from most general-purpose AI tools. Platforms built specifically for children, with age-appropriate design and safety guardrails built in from the ground up, are categorically different from adult platforms with age gates attached.

Pay attention to your state's legislation — and engage

The 134 bills moving through legislatures right now will shape what tools your child's school can use, what data those tools can collect, what rights you have as a parent, and whether your child graduates with AI literacy skills or without them. Most of these bills are moving with limited public engagement — because most parents don't know they exist.

Finding your state's tracker (FutureEd and MultiState both maintain current trackers that are publicly accessible) and understanding what your state is or isn't doing puts you in a position to engage meaningfully — at school board meetings, with legislators, with other parents.

Part Seven: What Educators Are Navigating Right Now

The legislative picture tells us what policymakers want. But educators — the teachers, principals, and curriculum directors who are actually implementing AI in classrooms — are living in a different reality.

Only 25% of educators worldwide feel they have been sufficiently trained to use AI effectively. Yet 95% of students and faculty are now using AI on campus daily. This gap between what educators are being asked to oversee and what they've been equipped to handle is not a minor inconvenience. It is a structural problem that most of the 134 bills either don't address or address with unfunded mandates.

A 2026 parent survey found that 78% of educators believe AI-assisted cheating has increased significantly since 2024, and 61% say they have caught students submitting AI-generated work at least once. Yet those same educators are working within institutional frameworks that often haven't given them clear guidance on what AI-assisted work even means — where the line is between AI as a tool and AI as a substitute, and how to evaluate student work in an era when the tools available have fundamentally changed what "original work" looks like.

Teachers are also navigating a specific tension that the research highlights: AI tutoring tools, when designed well, have real educational benefits. Khan Academy's Khanmigo, which launched as an experiment, is now integrated into thousands of school districts. Adaptive learning platforms that personalize instruction to an individual student's pace and gaps have demonstrated meaningful gains in certain subjects. The research on AI tutoring is not uniformly negative — far from it.

But the same research also shows that the benefits of AI tutoring are highly dependent on implementation. AI tutoring deployed thoughtfully, with educator oversight, aligned to curriculum goals, and used in ways that build rather than replace student thinking — this is different in kind, not just degree, from AI tutoring deployed as a shortcut, a way to reduce teacher contact hours, or a substitute for the kind of human relationship that the research consistently shows is foundational to learning.

Teachers know this distinction intuitively. They're living it. But without clear institutional frameworks — the kind that Oklahoma's SB 1734, Maryland's AI Ready Schools Act, and Idaho's SB 1227 are trying to create — individual teachers are left to navigate it alone, class by class, student by student.

The professional development gap is real. Maryland's legislation requiring compensated, statewide professional development through a train-the-trainer model addresses this directly. Hawaii's $5 million AI Education Grant Pilot Program does too. But these are outliers. Most states passing AI education legislation are creating obligations without creating the conditions for educators to meet them.

Part Eight: The Equity Dimension

The most underreported story in 2026's AI-in-education landscape is about equity. Specifically, about who benefits from AI in education — and who doesn't.

Research is beginning to confirm what was theoretically predictable: access to high-quality AI tools in education is not evenly distributed. The emerging AI divide is not simply about whether a student has internet access or a device. It's more granular than that, and more insidious.

According to 2026 research, the learning benefits of AI tutoring compound over time. Students who start developing AI fluency in elementary school build habits, vocabulary, and critical frameworks that give them a compounding advantage through high school and into the workforce. Students who don't have access to quality AI learning tools — or whose schools implement AI poorly, without training, oversight, or age-appropriate design — don't just miss the benefit. They potentially fall further behind as the gap widens.

The families with the most resources are best positioned to evaluate AI tools, supplement what schools provide, hire tutors who know how to use AI well, and ensure their children are building genuine competency rather than dependence. The families with the fewest resources are most dependent on schools to make those decisions well — and those schools are most likely to be the ones implementing AI without dedicated funding, without trained educators, and without the infrastructure to monitor outcomes.

One California study of adults' views on AI's impact on youth found that respondents "did not trust anyone — including the government, education systems or Big Tech — to regulate AI when it comes to their children." That distrust is not irrational. It reflects a history of technology companies entering schools with products that prioritized engagement over learning, data collection over protection, and scale over appropriateness. The AI wave is carrying the same dynamics — and the same communities that bore the brunt of previous ed-tech failures are positioned to bear it again.

The AI detection equity problem deserves specific mention. Studies have shown that AI detection tools disproportionately flag writing by non-native English speakers, students with certain neurodivergent characteristics, and students from demographic groups underrepresented in the training data of the detection models themselves. Without accuracy standards, without independent verification requirements, and without due process protections (which exist in law only in Florida), students from already marginalized groups face a disproportionate risk of academic consequences based on algorithmic outputs that have never been validated.

This is a civil rights dimension of the AI-in-education story that is getting far less attention than it deserves.

Conclusion: The Speed Isn't Slowing Down

134 bills in one legislative session. Student AI adoption jumping 26 percentage points in a single year. A landmark pediatric review in the journal Pediatrics. Federal legislation moving through Congress. International bodies issuing warnings. AI literacy about to be assessed on global standardized tests for the first time in 2029.

This is what a policy landscape looks like when it is racing to catch up with technology that moved too fast. And the honest assessment is that the gap between where the technology is and where the protections are is still very wide.

Some of these bills will pass. Some won't. Some will be strong; others will have loopholes large enough to drive a data center through. Implementation will be uneven. Funding will be inadequate. Educators will be asked to do things they haven't been trained for. And the technology itself will continue evolving faster than any legislative cycle can track.

But something meaningful is happening. The researchers, the pediatricians, the legislators, and the parents are all arriving at roughly the same conclusion at roughly the same time. The free-market approach to AI and children — release the product, see what happens, respond to the most visible harms — is not sufficient. Children are a special category of user. Their data deserves special protection. Their developing cognition deserves age-appropriate design. Their mental health deserves AI tools with actual safeguards. And their families deserve transparency about the tools their children are using.

AI literacy as a graduation requirement is not a threat to innovation. It's an acknowledgment that the children currently in K-12 classrooms will be the first generation to enter an adult workforce where AI fluency is a baseline expectation — not a specialization, not a competitive advantage, but a floor. How they get there — whether the tools they learn with are safe and appropriate for their developmental stage, whether the data they generate is protected, whether they build genuine cognitive capacity or outsource it — these are not abstract policy questions. They are questions about specific children, in specific households, in specific schools, right now.

The families paying attention now — the ones having the conversations, asking the questions, and engaging with what's happening in their state legislatures — will be the ones best positioned to navigate what comes next. That's not a small thing. In a space moving this fast, being informed is itself a form of protection.

At HeyOtto

We believe that the standards 2026's legislative wave is demanding of the AI industry are the standards every child deserves — not as minimum compliance thresholds, but as the foundation of how AI for children should be designed. Our KORA safety benchmark, COPPA-compliant architecture, age-differentiated content controls, and parent dashboard were built to reflect exactly that. We're proud to be ahead of where the law is going — and committed to staying there.