AI in Classrooms Faces Federal-State Clash Over Education Policy and Regulation

“You want federal law to create a floor, not a ceiling.”

Joseph South, Chief Innovation Officer at the Association for Supervision and Curriculum Development

Artificial intelligence is entering American classrooms faster than the policies designed to govern it. In the past year, state legislatures have moved to define how these systems should be used in schools, introducing measures that span AI literacy requirements, vendor transparency, student data protections, and limits on high-risk applications. At the same time, federal policymakers are signaling a different priority, emphasizing the need to limit regulatory fragmentation and preserve flexibility for continued AI development.

This moment brings two approaches into alignment. Education policy in the United States has long been shaped at the state level, where systems adapt to local needs and conditions. AI policy is increasingly framed through national priorities tied to economic competitiveness, security, and technological leadership. As these approaches meet in the classroom, authority and implementation come into sharper focus.

Meanwhile, state-led efforts reflect a model built on responsiveness and iteration. Federal signals point toward coordination and constraint. Each carries different assumptions about how policy should function in practice and how quickly it should adapt to new technologies.

Oversight is now being defined in real time as AI becomes embedded in classrooms. States continue to legislate within a system they have long governed, while federal policymakers signal a broader role in setting the terms. Variation across jurisdictions introduces uneven standards, while national approaches raise questions about flexibility and implementation. The path forward will shape how schools integrate AI and how they respond as new uses and risks emerge.

To better understand how this tension is unfolding in practice, we spoke with Joseph South, Chief Innovation Officer at the Association for Supervision and Curriculum Development (ISTE/ASCD) and former director of the Office of Educational Technology under the Obama administration. With experience shaping national education technology strategy, South offers perspective on how federal and state roles are evolving, and what is at stake as AI policy takes shape in classrooms.

Joseph South, chief innovation officer, is a strategic national educational technology leader focused on evidence-based learning transformation.

He formerly served as the director of the Office of Educational Technology at the U.S. Department of Education. In his role at the department, he was an adviser to the Secretary of Education and developed national educational technology policy, formed public-private partnerships to assist state and local education leaders in transitioning to digital learning, helped school districts expand the use of openly licensed educational resources (OERs), and collaborated with stakeholders to nurture a robust ecosystem of edtech entrepreneurs and innovators.

He also worked on a cross-governmental team to bring high-speed broadband, interactive devices, professional development for educators and leaders, and high-quality affordable digital content to U.S. classrooms. He is a strong proponent of learners’ active use of technology.

South has led learning product development teams at startups, museums, nonprofits, corporations, and higher education institutions. He has also directed a host of learning programs and consulted on projects in China, Korea, Mexico, South America and the Middle East.

He holds a doctorate in instructional psychology and technology from Brigham Young University.

The current tension over AI in classrooms is unfolding within a system that has long placed primary authority over education at the state level. From curriculum standards to classroom implementation, states have historically shaped how learning environments operate, with federal involvement focused on establishing baseline protections and providing support rather than directing policy.

“Education has been a state responsibility since the founding of the country,” South explains. 

His view reflects a widely understood structure in U.S. education governance, where federal engagement is most effective when it reinforces core protections rather than prescribing how systems should function. In practice, that role has centered on safeguarding civil rights and ensuring access, while leaving decisions about pedagogy, technology adoption, and classroom integration to state and local actors.

The emerging federal posture on AI reflects a different set of priorities. Policymakers are approaching the technology through a national lens shaped by economic competitiveness, security concerns, and the pace of global development. That framing places emphasis on maintaining flexibility for innovation and limiting constraints that could slow deployment. Within that context, education becomes one application among many rather than a domain requiring distinct governance considerations.

South notes that when federal policy aligns with its traditional role, it tends to support state action rather than constrain it. The concern arises when that balance shifts. 

“The role of the federal government in education is to protect the civil rights of educators and students and to support states in what they’re doing,” he explains. When policy moves beyond that scope, the fit between federal direction and state-led systems becomes less clear.

This mismatch is beginning to surface as AI moves from experimentation into daily classroom use. States are advancing policies that address immediate concerns about student data, vendor accountability, and appropriate use. Federal signals, by contrast, emphasize coherence and restraint across jurisdictions. These approaches operate on different timelines and assumptions, creating friction in how rules are defined and applied.

The issue reflects how governance is structured in practice, and where potential conflict emerges. A system designed for decentralized decision-making is now intersecting with a technology being shaped through national strategy. How those two models interact will determine how effectively policy can keep pace with classroom realities.

As AI systems evolve, the design of federal policy begins to determine how effectively schools can respond to new risks. The question is not simply whether regulation exists, but really how it is structured and how much flexibility it allows at the state level.

“You want federal law to create a floor, not a ceiling,” South explains.

The distinction shapes how governance functions in practice. A floor establishes minimum expectations for safety, privacy, and security, ensuring that all students are protected regardless of where they live. It leaves room for states to build on those standards as new use cases emerge. A ceiling, by contrast, limits how far states can go, constraining their ability to respond as technology evolves.

And that distinction carries practical implications in a classroom environment where AI capabilities are changing quickly. New applications are introduced, new behaviors emerge, and new forms of misuse follow.

“New use cases emerge, new threats emerge, and then states have their hands tied,” South notes.

A framework that restricts state action risks falling out of step with these developments, particularly when risks are not fully understood at the time policy is written.

Recent examples illustrate how quickly those gaps can appear. Tools that generate synthetic imagery have been adapted in ways that raise immediate concerns in school settings.

“Nudification apps and AI create child sexual abuse material,” South says, pointing to a category of harm that did not factor into earlier policy discussions.

These developments underscore how difficult it is to anticipate every use case in advance, and how quickly new risks can move from theoretical to urgent.

Federal policy still has a clear role to play in setting baseline expectations. Minimum standards for data protection, transparency, and appropriate use provide a foundation within which schools and vendors can operate. Existing frameworks, including laws governing student privacy, offer a starting point, but AI introduces new forms of data use and interaction that extend beyond those earlier models.

The challenge lies in how those standards are defined. A framework that establishes clear protections while preserving room for state-level adaptation allows policy to evolve alongside the technology. A framework that restricts flexibility risks slowing schools’ and policymakers’ ability to respond as conditions change. In an environment where both capabilities and risks are advancing quickly, the structure of regulation becomes as important as its substance.

As states move ahead with their own approaches, the practical effects of variation begin to surface. Differences in requirements across jurisdictions introduce complexity for school systems and for the companies developing and deploying AI tools. In areas such as student data governance, vendor obligations, and acceptable use, even small divergences can create operational challenges when systems are used across multiple states.

This dynamic is not new. Earlier efforts to regulate digital privacy in education followed a similar path, with states stepping in to fill gaps left at the federal level. The result was a landscape defined by overlapping and sometimes inconsistent rules. For companies operating nationally, that fragmentation carried real costs.

As South notes, “the money that we’re going to spend on in innovation, they instead spend on 50 versions of compliance.”
Resources that might otherwise support product development or safety improvements are redirected toward meeting a patchwork of requirements.

At the same time, uniformity alone does not resolve the underlying challenges. A single national approach can reduce complexity, but it also limits the ability to test different policy models and adapt to local conditions. Schools operate within varied environments, shaped by differences in infrastructure, staffing, and student needs. Policies that work in one context may not translate directly to another, particularly as AI tools are integrated into daily instruction.

These constraints are especially visible in how AI systems are deployed in classrooms. In some cases, students interact with general-purpose models that were not designed for educational settings or minors.

South points out the risks of this approach, noting that unrestricted access to frontier systems exposes students to capabilities and content that are not appropriate in a school environment. Purpose-built tools, by contrast, can incorporate safeguards aligned with existing expectations for educational technology.

The tradeoff is not between fragmentation and coordination as mutually exclusive choices. Each introduces its own constraints, and the balance between them shapes how innovation and safety evolve together. A system that leans too heavily toward fragmentation increases compliance burdens and reduces scalability. A system that leans too heavily toward uniformity narrows the space for adaptation. The policy challenge lies in structuring coordination to reduce unnecessary complexity without limiting the ability to respond as technology develops.

As the contours of this debate come into focus, a workable approach begins to take shape around how responsibilities are divided rather than consolidated.

Federal policy is positioned to establish baseline expectations, while states remain the primary site of implementation and iteration. The effectiveness of that arrangement depends on how clearly those roles are defined and how much flexibility is preserved over time.

At the federal level, the priority centers on setting minimum standards that apply across jurisdictions. These include protections around student data, transparency in how AI systems operate, security safeguards, and limits on inappropriate use. Establishing these baselines creates consistency where it is most needed, particularly in areas tied to student safety and rights. It also provides a reference point for schools and vendors operating across state lines.

Beyond those minimums, the system relies on state-level policymaking to adapt to changing conditions.

“They are more likely to get it right than Congress will get it right on day one,” South says, referring to the ability of states to experiment, learn from one another, and refine their approaches over time.

In practice, this creates a distributed process in which policies can be tested across different contexts and adjusted as new information becomes available.

This approach also reflects the pace at which AI systems are evolving. New capabilities and risks continue to emerge in ways that are difficult to anticipate at the national level. A framework that allows states to respond within defined guardrails creates room for adjustment without requiring constant federal intervention. It also supports a feedback loop in which lessons from state-level implementation can inform future national standards.

The role of educators and technology providers remains central within this structure. Schools are responsible for how these tools are introduced and used in practice, while developers shape the capabilities and constraints of the systems themselves. South states the need for both groups to engage directly with the technology and its implications, rather than relying on static assumptions about how it will function in the classroom.

Ultimately, a durable framework depends on aligning these layers. Federal standards define the boundaries, state policies shape how those boundaries are applied, and schools determine how AI is integrated into daily learning. Set the floor, not the ceiling. When those elements move in coordination, the system can adapt as the technology develops while maintaining consistent protections for students.