Screens Off, Questions On

Introduction

The K-12 education sector is navigating a broad backlash against the role of technology in children’s lives. The 2010s were full of promise, excitement, and hype that online platforms, apps, and one-to-one device programs could transform the classroom experience and bring new and more personalized educational opportunities to children. Smartphone and social media use among children exploded. And after schools closed off in-person instruction in early 2020 during the COVID-19 pandemic, one-to-one device programs became nearly ubiquitous.¹ Device use is almost universal in 2026, with 96% of teachers reporting that students use some form of digital device in their classes.² 

In recent years, however, the pendulum began to swing the other way: State attorneys general sued social media companies for their impact on the mental health of children;³ most states now restrict smartphone use in schools;⁴ and the rise of artificial intelligence (AI) is sparking new concerns about the role of technology in education, data privacy, and student well-being.

Building on successful campaigns to restrict cellphone use in schools, advocates are pressing policymakers to address broader issues related to technology and AI access in schools. There is growing momentum to regulate K-12 technology on two fronts: students’ screen time and use of AI. States such as Florida,⁵ Iowa,⁶ North Carolina,⁷ Ohio,⁸ Oklahoma,⁹ and Utah¹⁰ have passed laws or regulations on this front: Some set restrictions for districts to follow, while others require districts to set policies within a state framework. And across the country, school districts are moving ahead of state action. Notably, New York City Public Schools¹¹ and Los Angeles Unified School District¹² have set their own policies to limit screen time and AI use in classrooms. 

Regulating screen time and AI in schools is a more nuanced challenge than addressing student cellphones and brings a greater risk of unintended consequences. Few K-12 classrooms use cellphones as part of instruction. But district-purchased devices and software are a growing part of the curricula, interventions, and assessments that make up the core of instruction in schools — all while AI is increasingly being integrated into platforms and products schools already use. The status quo of tech use in most districts makes adding limits to screen time or AI use in schools a more complicated question than it may seem on the surface.

Parents, advocates, and policymakers are rightfully concerned about student well-being, data privacy, and the instructional quality of the technology students are using in schools. However, a desire to respond quickly to concerns about the use of technology in schools could lead to unintended consequences that disrupt classrooms, short-circuit promising innovation, and undermine strategies to improve instruction for a wide range of students. Instead, policymakers should start with a clear understanding of instructional strategy and align proper use of technology — including screen time and AI — with that strategy. Before voting on an ed tech regulation, policymakers should be able to answer three questions:

1. Does this regulation align with our instructional strategy and support student safety and well-being?

• • State legislatures and many districts across the country are focused on improving literacy and numeracy instruction. Will a regulation on screen time or AI tools align with ongoing efforts to improve instruction? 

2. What role is screen time or AI currently playing in schools?

• • Students and educators are surely using devices and AI tools both formally through district-procured platforms and informally outside classroom settings. Understanding the current state of play can help policymakers better tailor policy to the reality students and educators are experiencing. 

3. What replaces the products that are removed, and how will we know if the new approach works better?

• • Every meaningful restriction on screen time or AI tools creates an opening for students and educators. What will fill that void? Are there plans to support students and educators through that transition? And most importantly, how will we measure whether the new approach to technology leads to better results for students?

There are different roles for state policymakers and district leaders to play in navigating this terrain. State policy can set a common floor of expectations and provide districts with support and incentives. Decisions about how technology best fits in classrooms should be made by local school systems that have to navigate the practical realities of implementation. And both state and district policymakers can work together to ensure that unsafe and ineffective tools are kept out of classrooms (Table).

TABLE: COMMON STATE APPROACHES TO STUDENT AND TEACHER SCREEN TIME AND AI USE IN K-12 SCHOOLS

This analysis amplifies key questions state policymakers and district leaders should ask before making final decisions about the role of screen time and AI in classrooms so they do not unintentionally ban a product that is creating great value for a group of students or teachers. It also illustrates how policymakers can draft rules along these five criteria — Exposure and Access; AI’s Role and Capabilities; Permission and Supervision; Misuse and Consequences; and Evidence and Impact — to shape the role technology plays in instruction. It also details how these policy levers work and common pitfalls to avoid.

Exposure and Access

Exposure and Access rules control how much, what type, when, and in which grade levels students use devices or products that include AI. Limitations on access to screens and to AI products are the core of this wave of policymaking.

Policy Mechanics

Key policy elements in this category include: 

1. Global restrictions: Impose age-based bans or usage limitations. 
2. Hardware restrictions: Shift from one-to-one devices to a shared or “cart” model.
3. Software restrictions: Prohibit particular types of ed tech or AI software products. 

Global restrictions on student screen time and AI use are often scaled by age, placing the tightest restrictions on those technologies in the early grade levels and permitting more technology use by high school students. A common approach is a total or near-total ban on screens for young students, typically from pre-K through Grade 2, and a limit on the number of minutes that older elementary students are allowed to use screens for learning, while AI restrictions often extend into middle school. High school students face the lowest level of restriction on screen time and AI use, so policies can ensure some degree of exposure to and fluency with technologies that are widespread in postsecondary and workforce settings.

Some technology restrictions are more granular, focusing on specific uses of hardware or software. On the hardware side, there is a growing shift from one-to-one device programs to a shared cart model to reduce student use of Chromebooks or tablets. Other policies focus on restricting which software students can access, from limitations on consumer software (e.g., YouTube, ChatGPT) to restrictions on education-specific tools: In Florida, Hillsborough County Public Schools ended the use of Amira, i-Ready, and Imagine Learning in “daily classroom instruction” for grades K-2.¹³ 

These approaches to restricting students’ technology use are often accompanied by exemptions. These can include exemptions for students with learning differences and multilingual learners who may rely on specialized software and hardware to meet their learning needs. District- or state-mandated testing is another use case typically exempt from technology restrictions.

Potential Pain Points

• Clearly defining screen time and AI: The precision with which “screen time,” “AI,” or other ed tech terms are defined shapes how policies will affect classrooms. For example, a recent law in Iowa regulates “lessons, assignments, assessments, or instructional activities delivered through instructional technology,”¹⁴ while Utah’s law regulating screen time defines it more narrowly: “Screen-time” counts only device use that “does not involve instruction, guidance, or interaction” with a teacher or other educator.¹⁵ 
• Clarity on AI regulations: AI restrictions can be narrowly tailored to address student use of generative AI tools (e.g., ChatGPT, Claude, Gemini) or broadly written to cover all products with generative AI features. The latter could become more difficult to enforce as ed tech and consumer tools integrate generative AI features into their core products.
• Implementation oversight: Of all the policy elements in this conversation, bans and limitations are likely the most straightforward to adopt. Even so, implementation can quickly become unwieldy. If students get an hour of total screen time per day, what determines which teachers or subjects “use” those minutes? How will anyone know whether a group of students goes over time one day, or most days? If the policy includes limits on how much time students can spend on screens doing homework, how will that be implemented and monitored?
• Student preparation for an AI-driven future: This category of policy — Exposure and Access — should also include the work that education systems owe students to prepare them for a workforce being transformed by AI. States such as Maryland and North Carolina are adding AI literacy standards, but district policies do not always require AI literacy education. The challenge for policymakers is to ensure that restrictions on screens and AI align with efforts to prepare students for postsecondary and workforce environments that are likely to be heavily infused with those technologies.

AI’s Role and Capabilities

Rules governing AI’s Role and Capabilities control what the AI tools may do, what data they may use, and how they must behave with students. As AI becomes more deeply intertwined with products that school systems already use, states and districts are looking to restrict how students and educators use AI.

Policy Mechanics

Key policy elements in this category include:

1. Limits on AI’s role in official decisions: Often related to consequential decisions such as grading, discipline, placement, or promotion (for students); and performance ratings and promotion (for teachers).
2. Limits on data use: Restrictions on what student data an AI system may use, collect, or act on.
3. Behavioral design restrictions: Restrictions on how AI may interact with students, often prohibiting AI that is designed to elicit an addictive response or simulate an emotional relationship with a student.
4. Crisis-response requirements: Obligations that an AI system must take certain actions if it detects signs of self-harm or suicidal ideation (such as referring the student to a human or a crisis resource rather than responding on its own).

The clearest and most common element is a limit on the use of AI in high-stakes decision-making. Oklahoma bans AI from being the “primary basis” for a student’s grade, discipline, placement, or promotion;¹⁶ in Massachusetts, Boston Public Schools uses a narrower “sole basis” standard for the use of AI in grading, discipline, and general academic evaluation.¹⁷ The difference in wording matters: A “sole basis” standard can be satisfied by a nominal human sign-off even if that review is cursory, while a “primary basis” standard establishes more of a gray area about how much weight the AI’s output carried in the outcome. 

Data use and behavioral design restrictions are newer and so far concentrated in a handful of states. The Florida State Board of Education’s adopted AI policy bans any AI instructional tool from engaging in undisclosed behavioral monitoring, social scoring, or psychological profiling, and separately bans tools “designed, marketed, or configured” to meet students’ social or emotional needs, regardless of parental consent.¹⁸

A related and rapidly expanding body of law addresses behavioral design restrictions and crisis response. California,¹⁹ Idaho,²⁰ Oregon,²¹ and Washington²² have all enacted laws requiring chatbot operators to disclose to users that they are interacting with AI — with extra notifications for minors in some of these states — and to respond to expressions of self-harm or suicidal ideation and route the user to crisis-response resources. These are not education-specific laws; they are consumer-protection statutes mostly aimed at companion-style chatbots, and some states (Oregon and Washington) exempt educational tools. As a result, districts cannot assume the classroom tools they procure are covered. 

Potential Pain Points

• High-stakes decision-making enforcement: The “primary basis” versus “sole basis” distinction for high-stakes decision-making is harder to enforce than it looks on paper. Proving how much an AI output drove a grading or discipline decision requires reconstructing a teacher’s reasoning after the fact — there is no clean way to audit whether a policy like this is being followed in practice, as opposed to being satisfied by a pro forma human signature. 
• Oversight of student data privacy: Data use restrictions carry a similar problem. “Undisclosed behavioral monitoring” and “psychological profiling” are not self-evident categories, and a district has limited technical capacity to verify what a vendor’s model is doing with student data behind an application programming interface. This is largely a vendor compliance question dressed up as a district policy, which means the real enforcement point is procurement contract language, not classroom practice.
• Mapping disclosure and crisis-response protocols: The companion chatbot laws were written for consumer products, not classroom tools, and they draw the line in different places. They can exempt software used solely for education (Oregon) or exempt narrowly tailored, curriculum-aligned tools that do not offer “open-ended conversational companionship” (Washington). A district adopting an AI tutor may assume the vendor follows these laws when the tool is legally outside them, or when it is covered in one state but not the next. If a district wants disclosure and crisis protocols, it should write them into the contract rather than rely on the statutes.

Permission and Supervision

Permission and Supervision rules clarify and tighten regulations on who is authorized to make decisions about which specific technology or AI product students and teachers use, when they use it, and how that use is supervised.

Policy Mechanics

Key policy elements in this category include: 

1. Tool and vendor approval or certification lists.
2. Parent and caregiver authority to opt in to or opt out of certain programs, and related disclosure requirements.
3. Institutional oversight and educator-directed supervision requirements.
4. Time and place restrictions.
5. Policies on how students may use AI supports to complete work, particularly homework.

Permission and Supervision policies allocate authority among four decision-makers: states, which set boundaries and establish standards; districts, which vet tools before they are used in classrooms; teachers, who grant permission for a specific use in the moment; and parents, who may have the option to opt a child in or out of a district-approved tool. States are increasingly setting requirements for districts to adopt policies, with some shaping the role of teachers and parents to influence technology use in schools. Historically, districts have generally taken a “default-allow” approach to web and application access: permitting open access while prohibiting specific restricted content. However, there is an increasing move toward a “default-deny” approach: permitting access only to preapproved applications. For example, Greenville County Schools in South Carolina now limits students to applications approved by the district.²³ In Virginia, Fairfax County Public Schools will pilot a default-deny approach in six schools.²⁴

Parental authority varies in both direction and substance. Iowa’s policy allows parents to request reductions in digital instruction beyond what state law requires.²⁵ Florida’s new policy runs the other way, requiring affirmative parental opt-in before a child directly uses AI instructional tools, while banning companion-style AI outright, regardless of consent.²⁶ These defaults will be important as districts begin to implement the policies: Opt-in and opt-out approaches to the same tool will likely produce different usage rates depending on whether parents must take action to allow or restrict access. Technology opt-out policies may also prove more difficult for educators to implement, given the widespread use of digital curricula in classrooms and the complexity of managing students with differing levels of parental consent.

Supervision requirements govern how closely an adult is involved once a tool is approved. Oklahoma requires all classroom AI tools to be “educator-directed,” meaning under the supervision and professional judgment of a teacher or school employee.²⁷ In Florida, Orange County Public Schools requires that students use AI for assignments only in accordance with class guidelines and with prior documented approval,²⁸ but teachers will not be able to rely on software to detect AI use,²⁹ leaving enforcement to teacher judgment. And as approved tools add new features (which they do frequently), district leaders will need a process to determine whether those updates remain aligned with district policies.

Potential Pain Points

• Student use of AI on coursework and homework: Using AI to complete coursework and homework is a significant issue that districts have been grappling with since the launch of ChatGPT. Districts urgently need to set and clarify guidelines so that students and families know what is allowable and what is not. This is a difficult policy element, since classroom-based supervision does not extend to the home, and products that detect AI use are not completely reliable.  
• Clarity in what opt-in and opt-out entail: The specific wording in the policy matters as much as it does for Exposure and Access rules. Opt-in and opt-out will be very different to administer and will likely yield significantly different usage rates, with opt-in policies generally resulting in lower use. Both policies create a practical challenge for educators, who would need to manage two cohorts of students: those who are able to use a particular technology and those who are not. If those challenges are not addressed directly, an opt-out or opt-in policy on paper could become a de facto ban on a particular tool or technology. 
• Clarity in what supervision means: A policy can define “supervision” as a teacher’s physical presence, approval obtained in advance, or review of output after the fact — and these are not interchangeable. Permission rules are also inherently harder to enforce consistently than an outright ban, since they depend on a judgment call made repeatedly by different adults rather than on a single yes-or-no condition.
• Quality control for alternative options: As with Exposure and Access policies, opt-in and opt-out policies will require that a high-quality alternative be in place to fill in for the ed tech product that students will not be using. Unlike in Exposure and Access policies, these rules create two different instructional programs running in parallel — one with ed tech products and one without — which complicates implementation with fidelity.
• Reviewing updates to approved tools: Ed tech products frequently roll out new features through software updates. If approval processes do not account for the evolution of products post-approval, tools may add features that conflict with district guidelines.

Misuse and Consequences

Misuse and Consequences policies control what happens when screen time and AI rules are violated. Routine AI and technology governance (acceptable use policies, academic integrity codes, staff conduct guidance) overlaps with this area of policymaking. Districts are moving to adopt formal AI policies, but a scan of 219 districts in 43 states by the Center for Education Market Dynamics found that 44% “have no formal AI policy adopted or in progress.”³⁰ As new state and district policies regulate screen time and use of AI, what happens when someone breaks the rules established through those policies? This is the most underdeveloped element of this current wave of policymaking, with very few examples of clearly defined consequences across district- and state-level policies.

Few of these policies define new consequences for student misuse of AI or violations of screen time or other technology restrictions; most point to existing academic integrity or student conduct codes. By design, this wave of policymaking seeks to hold adults in the system accountable for establishing strong controls around AI and screen usage with students, presumably leaving student consequences for misuse to be defined in existing district policy.

Policy Mechanics

Key policy elements in this category include:

1. Educator- or staff-facing consequences. 
2. District or school consequences.
3. Vendor consequences.

Educator- or staff-facing consequences for violating screen time or AI use rules are not common in recently approved policies. One of the clearest examples is Alabama’s House Bill (HB) 78:³¹ In childcare, pre-K, and kindergarten settings, the first response to noncompliance with the law’s screen time limits is additional training. If problems continue, consequences vary by setting: In childcare and pre-K, the state may place programs on a corrective action plan, while kindergarten teachers’ continued noncompliance is an ethics violation that can lead to discipline. That distinction (taking action on a program versus on an individual educator) is a policy choice about whether noncompliance is treated as an organizational issue or a professional conduct issue.

Consequences for districts and schools that do not comply with these policies are also rare in the recent round of policymaking at the state and district levels. Utah’s HB 273 allows the state board to withhold or delay a district’s ed tech funding until it demonstrates compliance.³²

Policies addressing vendor obligations are nascent — the most stringent consequences for AI vendors currently fall on the consumer product side, where laws often include exceptions for education-focused products. In California, people harmed when a chatbot operator fails to meet its disclosure and crisis-response requirements can sue for damages,³³ while New York allows its attorney general to seek civil penalties.³⁴ These policies often draw on of a more mature regulatory lane through which vendor noncompliance terms have been negotiated for years, but it will be critical to update and adapt these policies for the speed of change and complexity of AI.

Potential Pain Points

• Overreliance on outdated or incomplete academic integrity codes: Pushing student consequences onto existing academic integrity codes assumes those codes have been updated to handle the new realities of AI — a faulty assumption, given that many districts are scrambling to catch up with students’ use of AI. 
• Lack of clarity and subjective application of consequences: The lack of clear processes for determining misuse and of clear consequences for staff, schools, and districts may invite uneven and inequitable application of these new policies. This could also undermine the intent of the policies if staff, schools, or districts perceive that there are no consequences. Vendors of consumer AI products face the most clearly specified consequences of anyone touched by this policy wave, but again, carve-outs can exempt ed tech products. Absent clearly articulated consequences for noncompliance, new guidelines and regulations on ed tech will introduce uncertainty for students, educators, and administrators to navigate.

Evidence and Impact

Policies addressing Evidence and Impact control which products are eligible for use, and how future evidence of impact on student learning and well-being is created. Parents and policymakers want to know whether the screens and AI products used in schools are improving learning outcomes. Such evidence is limited, especially for products that include AI, because they are so new and changing so quickly. Stanford University’s Systems Change Advancing Learning and Equity Initiative review found that only 20 of more than 800 academic papers on AI in K-12 education met the bar for high-quality causal evidence, and none of those studied U.S. student AI use directly.³⁵ If this round of policymaking does not lead to more evidence on how screens and AI affect student learning, it will be a significant missed opportunity.

Policy Mechanics

Key policy elements in this category include:

1. Preadoption efficacy requirements: Establish evidence that a tool must demonstrate before a district may adopt it, rather than evidence collected afterward.
2. Diagnostic or baseline research: Study the problem to inform a future decision.
3. Compliance monitoring: Verify that an existing rule is being followed.
4. Impact or outcomes evaluation: Measure whether a policy produced the desired outcome.

Guidance requiring evidence of effectiveness before a tool is adopted is mostly aspirational. Florida comes closest: Its State Board of Education’s new rules require vendors to disclose any independent studies of a tool’s effectiveness,³⁶ but the absence of a study cannot by itself block approval. Maryland requires districts to vet AI tools with a state rubric,³⁷ but the rubric evaluates mainly for privacy, alignment, and consistency with state guidance rather than for evidence that a tool improves learning.

Diagnostic or baseline research requirements are more common in both state and district policy. There is little data on screen time, AI, and broader ed tech use in K-12 schools; some new district policies are seeking to close that evidence gap. For example, in Maryland’s Montgomery County Public Schools, its County Board of Education requires a K-8 tech audit and report by March 2027.³⁸ 

Compliance monitoring and impact or outcomes evaluation include few requirements in current state and district policies to verify compliance or noncompliance, and few requirements that the impact of these new policies be measured and evaluated. This mirrors findings in the Misuse and Consequences category: Just as few policies in this wave specify what happens when a rule is broken, fewer still specify how anyone will know whether a new rule worked. Policymakers have been far more active in defining restrictions than in verifying either compliance with them or their effects.

Potential Pain Points

• Preadoption evidence requirements: The biggest challenge in this category concerns pre-adoption evidence requirements, which are a reasonable response to a sector awash in low-quality ed tech applications. However, if those requirements are too rigid, they could exclude nearly every AI-powered product on the market today and many other ed tech products rather than limit access to the weak ones. While well-intentioned, this policy approach could close off avenues for potentially significant positive impact on classroom instruction and student outcomes. 
• Ed tech research for research’s sake: There is also a risk that diagnostic research on the current use of ed tech and AI in schools is seen as an end unto itself. Such research should serve as a baseline to inform changes to screen time, AI, and other ed tech policies — that connection should be clearly established as the intent of that research. The same holds true for evaluations of screen time and AI compliance or impact: A report that only exists as a PDF on a website does little to inform policy. Instead, state policymakers and district leaders should draw clear lines to establish who reviews and acts on the results of usage, compliance, or effectiveness data collection. 

Conclusion

As the pace of technological innovation accelerates, states and districts are moving to address how technology is used in schools. There is a risk that, as state policymakers and district leaders try to move quickly to regulate these products, there will be unintended consequences. Focusing on specific technological questions can obscure practical impacts on teaching and learning. It is easy to design policies that restrict tools; it is an entirely different challenge to craft policies that work toward a coherent role for technology in classrooms. 

As state policymakers and district leaders revisit how technology is used in schools, they should keep these questions front and center:

1. Does this regulation align with our instructional strategy and support student safety and well-being? Students learn when they work through problems that push their thinking. Good technology policy keeps students from outsourcing cognitive effort and gives teachers access to tools that help them deliver high-quality, differentiated instruction. Policies restricting technology use should ensure that changes in how classrooms operate align with existing strategies to improve instructional quality and outcomes for all student groups.
2. What role is screen time or AI currently playing in schools? Screen time and AI use are part of almost every school district, whether through official, district-procured tools or through unofficial use by students and educators outside official channels. Policies that seek to shift this dynamic should begin with an understanding of the current state of technology use in schools. 
3. What replaces the products that are removed, and how will we know if the new approach works better? Policymakers drafting regulations need to think through the secondary and tertiary impacts of their decisions. If classroom technologies are restricted, what will students and teachers do instead? Will the new approach be better than what it replaced, and for whom? Who will be responsible for making this determination? 

Powering down screens and blocking AI tools are straightforward but insufficient steps for responsible policymakers. New rules governing the use of technology in schools can lead to stronger, more coherent instruction in classrooms or disrupt it. The questions highlighted in this analysis can help state policymakers and district leaders pressure-test proposals they will see this year and in the years ahead as these technologies continue to evolve.