Introduction
The rapid implementation of artificial intelligence (AI) into social, economic, and political spheres has created a need to determine the best practices to develop comprehensive AI literacy. Schools are actively taking part in technological advancement, and world leaders are considering compulsory education as a solution to AI illiteracy, as younger generations will enter an increasingly algorithmic world. Many educators are most concerned with the curricula and regulatory frameworks that would allow students to develop technical and ethical understanding, both of which are important for students’ engagement with emerging technologies. The rapid implementation of AI tools has also led educators to question whether students can interpret and evaluate new information they see online. Amid discussion of AI literacy, it is necessary to also consider the unequal distribution of technological resources within the United States to reimagine how policy implementation can shape the future.
Marginalized communities within the U.S. face various challenges that limit them from succeeding in the educational system, many of which stem from social and cultural factors. Most importantly, marginalized communities often have underfunded schools that don’t have the resources to adopt up-to-date technologies as fast as other schools can.[1] Introducing an AI curriculum without understanding structural inequalities would only further exacerbate these issues. Thus, we cannot simply allude to AI literacy policies as the solution to baseline access. The divergence of access not only contributes to the widening of individual economic mobility gaps; it also undermines democratic participation, as AI now mediates access to information, political expression, and economic security. In this context, policymakers should not rely on simplistic solutions or assume that introducing new frameworks alone will resolve existing inequities. Instead, AI literacy policies must be developed in ways that acknowledge the U.S.’ deeply rooted structural inadequacies. Only by doing so can we prevent AI literacy from reinforcing historical patterns of exclusion and ensure that it becomes a pathway toward educational and civic empowerment.
This paper argues that, if implemented without structural reform, AI literacy will reproduce the racial and socioeconomic inequities embedded in U.S. educational and legal systems, and therefore requires a federally guided, equity-centered governance framework. This paper proceeds in four parts. Section I situates AI literacy within the historical and structural inequalities of U.S. education policy. Section II analyzes the current legal and policy landscape governing AI, digital learning, and civil rights. Section III diagnoses how existing governance structures risk reproducing inequality when applied to AI literacy. Lastly, Section IV proposes a federal, equity-centered framework for AI literacy that integrates civil rights enforcement, targeted infrastructure investments, procurement reform, and community-based governance.
I. Historical Context: Educational Inequality and the Digital Divide
The unequal access to education, which is now shaping the decision for AI literacy today, is rooted in the political, legal, and economic decisions that have contributed to the structure of educational opportunity within the U.S. Education disparities did not arise from the simple construct of schooling itself; they were actively reconstructed through public policies that distributed resources unevenly across communities, and these policies reshaped educational opportunity to fit particular political and economic priorities instead of promoting equal access. Federal housing and education policy played a central role in this process, as the use of redlining practices denied mortgages to Black families and other marginalized groups; this placed white families into well-resourced suburbs and confined Black and low-income families to underfunded urban neighborhoods.
Before 1954 and, to some extent, in the following years, we can study history to conceptualize how federal, state, and local governments have contributed to educational segregation through both housing and school finance policies. Federal housing policy decisions in America have restricted Black families from gaining access to homeownership, which has led them to live in underfunded districts. The federal government created the Home Owners’ Loan Corporation’s redlining system in 1933, which color coded maps that labeled Black neighborhoods as “hazardous” and ineligible for mortgage insurance.[2] The creation of the Federal Housing Administration’s discriminatory mortgage insurance program (1934–1960s), another program implemented by the federal government, aimed to subsidize low-interest loans for white families while denying them to Black families.[3] This was an act of suburbanization that laid the foundations for well-funded and underfunded schools, thus deepening educational segregation through federal housing policy. Local governments also put forth zoning laws that prohibited multifamily housing, a policy specifically designed to keep Black families out of white neighborhoods.[4] Because they were ultimately preventing people from living in affluent neighborhoods, they were also protecting these communities from sharing valuable resources.
At the same time, state and local governments used school boundary drawing, exclusionary zoning, and suburban incorporation to preserve local tax bases, which protected wealthier districts from resource sharing and kept poorer districts segregated and underfunded. State funding formulas rarely corrected these disparities, and many plans favored communities with higher property wealth, which reinforced racial and economic inequalities instead of addressing them. These public policies created an educational landscape in which technological access and school resources flowed toward communities that already held racial and economic privilege, while underserved communities were left without the infrastructure needed to benefit from new forms of learning, now including AI-powered education. These disparities were constructed through public policies that distributed resources unevenly across communities, contributing to racial and economic segregation and granting technological access to already privileged groups while neglecting those in need. Acknowledging this history is vital because any AI literacy initiatives introduced in today’s unequal political infrastructure will further contribute to inequity unless policymakers undertake critical work to reverse these societal harms. Due to past policy decisions, AI should be viewed conceptually, with an understanding that addressing its implications in marginalized communities requires a holistic approach that tackles the harmful effects of redlining.
In Milliken v. Bradley (1974), the Supreme Court held that federal courts could not require interdistrict (across city–suburb) desegregation plans unless the suburban districts were proven to have intentionally contributed to segregation.[5] Because Detroit’s surrounding white suburbs had not been found liable, the Court ruled that they could not participate in metropolitan-wide desegregation.[6] The ruling effectively stopped white suburbs from integrating, even where urban segregation was clearly rooted in state-sanctioned housing discrimination, local zoning laws, and school district boundary manipulation. This political decision contributed to racial inequality and unequal access to public resources. The state consistently relies on property taxes to fund education, which contributes to wealthy districts being able to receive more funding for teachers, technology, and infrastructure than low-income districts. Thus, AI becomes a tool implemented through federal and state funding, which may have significant implications for minority communities, as funding is not always distributed equitably.
Because public schools rely heavily on local property taxes, these housing decisions produced districts with dramatically different funding capacities, which allowed suburban communities to build stronger schools with better teachers and more updated technology, while urban districts faced chronic underinvestment. This largely explains why poor districts and schools are limited in adopting best technological practices such as AI. These policies have further widened the gap in schools’ capacity to adopt advanced technology. Those same capacities continue to guide who gains access to AI tools, curriculum, and the necessary training. Technological advancements come after these structural inequalities, but they ultimately contribute to the foundational elements of the digital divide, which started to form in the late twentieth century. Federal policy initiatives such as the 1996 Telecommunications Act expanded internet access by restructuring the nation’s telecommunications system and creating federal programs to support school and library connectivity.[7] The Act established the E-Rate program, which subsidized the cost of broadband, telecommunications services, and internal network infrastructure for K–12 schools and public libraries.[8] By requiring telecommunications companies to contribute to a Universal Service Fund, the policy aimed to ensure that even low-income or rural districts could obtain discounted internet services.
In practice, the Act marked the first major federal effort to integrate digital connectivity into public education, laying the foundation for widespread classroom internet access, but it did not eliminate problems related to broadband inclusion, device availability, and technological training. Because of the rise of technological innovation, well-resourced schools started to integrate STEM programs, while underfunded schools struggled to get funding. Both “Education to Innovate” and “Race to the Top” were designed to increase STEM readiness, stimulate technological capacity, and help low-income districts adopt modern instructional tools.[9] But these initiatives were limited in their ability to change the harmful conditions in marginalized communities because they relied on competitive grant models rather than guaranteed federal investment.
Race to the Top required states and districts to demonstrate existing capacity, such as strong data systems, effective teacher pipelines, and modern technological infrastructure, to qualify for funding, but these were the very resources that underfunded districts lacked due to decades of property tax disparities, redlining, and state disinvestment.[10] As a result, the districts most in need of support were often the least able to compete for it. Neither program addressed foundational inequities such as broadband quality, device access, class size, teacher shortages, or the absence of equitable school finance structures. They encouraged innovation without supplying the long-term funding necessary to maintain it, and they assumed a baseline technological readiness that simply did not exist in many high-poverty districts. These programs helped well-resourced districts accelerate innovation while leaving structurally marginalized communities with the same barriers they had before: unequal infrastructure, limited digital capacity, and underfunding. Due to the political implementation of these policies, we have seen, and continue to witness, the consequential nature of those decisions, which have shaped access to housing, school funding, and digital infrastructure. Federal policies like HOLC and FHA lending systematically removed Black families from the wealth-generating opportunities of homeownership, creating tax-poor districts that still struggle to fund modern technologies. State policies that rely on property taxes to finance education reinforced these disparities, ensuring that underfunded districts could not invest in STEM programs or digital learning initiatives at the same pace as wealthier communities.
Local exclusionary zoning and rigid school district boundaries further concentrated poverty and racial segregation, which limited the technological capacity of marginalized communities for decades. Even federal digital policies, such as the 1996 Telecommunications Act and later competitive-grant initiatives like Race to the Top, expanded access unevenly because they assumed existing infrastructure that many under-resourced districts lacked. As a result, today’s disparities in AI literacy reflect the cumulative effect of these historical choices: students in affluent districts benefit from early exposure to devices, broadband, STEM courses, and AI-enabled tools, while students in underfunded districts continue to face inadequate connectivity, outdated technology, and limited curricular access. These are not new inequities; they are the modern manifestations of long-standing political decisions that distributed resources along racial and economic lines.
Guan et al. (2025) demonstrate this clearly: students whose parents hold a bachelor’s degree scored, on average, 26-30% higher on AI literacy assessments than students whose parents had a high school diploma or less, and students from high-SES families reported nearly double the frequency of hands-on AI tool use compared to low-SES students.[11]
Higher education has been impacted by similar patterns of inequality, as institutions with greater funding can adopt AI tools, develop specialized curricula to address the critical challenges within the world, especially the rise of AI, while students who attend institutions that don’t have access to funding face inconsistent or inequitable access. As Shoval (2025) explained, AI integration in universities can widen rather than narrow equity gaps when adoption is not grounded in structural accountability.[12] By rapidly outputting policy that aims to advance technological adoption and innovation, we risk exacerbating the existing inequalities without prioritizing the existing inequitable practices, which correlates to policy failure.
Federal and state educational policies have always been rooted in innovation without recognizing underserved communities and their voices within the framework. Policy decisions such as No Child Left Behind (2001) and the Every Student Succeeds Act (2015) are often framed as major federal interventions designed to improve educational outcomes, but their actual impact reveals how policy can overlook the structural inequalities that shape learning conditions. NCLB attempted to create national accountability through high-stakes standardized testing, requiring schools to demonstrate “adequate yearly progress” or face federal sanctions.[13] While the law positioned itself as a tool for equity, it operated on an assumption that all schools, regardless of race, income, or local tax base, had equal capacity to meet rising performance expectations. In reality, NCLB intensified pressure on already underfunded districts by tying success to test scores without addressing the resource disparities that made meaningful improvement nearly impossible. As a result, schools serving low-income and minority communities were more frequently labeled as “failing,” reinforcing stigma rather than supplying the investments needed to close achievement gaps. This is a prime example of what happens when we fail to address the historical problems that have contributed to minority disadvantage; instead, we design new metrics to quantify and explain why underserved communities are not meeting established standards. This could, in fact, also occur with AI if we do not address the root causes of unequal technological access. ESSA, signed in 2015 as NCLB’s replacement, promised to correct some of these shortcomings by shifting power back to the states and offering more flexibility in how schools are evaluated. However, ESSA preserved the central logic of accountability without resolving the underlying inequities that shape educational access. Although the law allows federal funding to support effective instructional technology,[14] it does not require states to confront disparities in broadband infrastructure, digital literacy, or access to advanced technology, issues that directly influence whether students can benefit from emerging AI-powered educational tools. ESSA therefore continues a long-standing pattern in federal education policy: prioritizing performance metrics and the rhetoric of innovation while leaving the deepest structural barriers, property-tax-based funding, racial segregation, and unequal technological capacity, largely untouched. Both NCLB and ESSA allow us to understand the limitations of reform efforts that center on assessment rather than equity. They demonstrate how federal policy can demand academic progress from marginalized communities without addressing the political decisions that produced unequal conditions in the first place. Also, the rapid implementation of online learning and digital learning strategies during the COVID-19 pandemic exposed the disparities within underserved communities, which disproportionately affected low-income students and students of color.[15]
The political history and implementation of digital technology are deeply important, as they allow us to understand the legal and political frameworks into which AI literacy must be introduced. As Singh et al. explain, marked disparities in computer and broadband use persist across racial/ethnic and socioeconomic lines in the United States, with access differing sharply not only by income but also by geography and educational attainment.[16] Research consistently shows that gaps in digital access remain profound: for example, only 66.0% of American Indian/Alaska Native adults have broadband access, compared to 89.7% of Asian/Pacific Islander adults and 83.5% of non-Hispanic Whites, while individuals living below the poverty line experience an 18-percentage-point deficit in computer and internet use relative to higher-income groups.[17] 43% of adults in households earning under $30,000 lack home broadband, 41% do not have a desktop or laptop computer, and 27% rely solely on a smartphone for internet connectivity.[18] Furthermore, Americans with lower incomes continue to be less likely than those with higher incomes to have home broadband, a smartphone, or a computer, underscoring that essential forms of digital access remain stratified along economic lines.[19] These disparities make clear that equal access to AI-enhanced learning environments cannot exist when entire communities still lack consistent broadband, adequate devices, or the digital capacity necessary to meaningfully engage with emerging technologies. This suggests that AI literacy frameworks should not focus solely on individual competence or compliance with established standards, but also on the structural policies that shape who can access, and ultimately benefit from, AI education. Educational inequity shapes how people engage in society, as it may relate to their economic mobility, political voice, and democratic participation. As AI becomes deeply embedded in society, it has already begun to influence housing outcomes, employment, resource allocation, social services, and democratic engagement.[20] In the housing sector, for example, AI-powered tenant-screening algorithms have reproduced existing racial and socioeconomic biases by disproportionately flagging Black and low-income applicants as “high-risk,” even when controlling for credit history or rental records.[21] Similarly, predictive policing technologies, often used in proximity to housing authorities, have concentrated surveillance in predominantly Black and Latino neighborhoods, reinforcing residential segregation patterns rather than alleviating them.[22] Therefore, equal access is a civil rights concern.
II. Legal and Policy Landscape for AI Literacy
The regulation of AI literacy is deeply interwoven with education law, civil rights protections, federal and state digital learning policies, and the establishment of rules for educational technology and AI governance frameworks. Many policymakers have been central to the idea of the U.S. being prepared for AI which, in addition to technical familiarity with emerging tools, requires a society capable of navigating the political, economic, and civil rights consequences that accompany algorithmic systems. Preparation, in this sense, means developing the legal, educational, and infrastructural capacity to ensure that AI strengthens democratic participation rather than undermines it. It entails building an equitable technological foundation, broadband access, device availability, curricular integration, and strong civil rights protections, so that communities are not excluded from the benefits of innovation. Most importantly, being prepared for AI requires a historically informed understanding of how past policies have reproduced inequality. Without confronting these structural conditions, policies aimed at “AI readiness” risk functioning as another mechanism that privileges already-advantaged districts while marginalizing those without the resources to engage. In this context, true preparedness is not simply about adopting new technologies; it is about constructing an equitable framework that allows all students to meaningfully participate in an AI-driven society. But within the true political dynamics, the policy landscape itself is deeply fragmented, especially when it comes to addressing the inequities that contribute to unequal access to education in the U.S.
A. Education Law and Federal Statutory Frameworks
Federal education law sets the foundational principles for states and districts, but it does not explicitly state what those obligations are when it comes to AI literacy. The Elementary and Secondary Education Act (ESEA) was put into place in 1965 and reauthorized by the Every Student Succeeds Act (ESSA), which denotes that educational accountability is a process that is built on standardized assessments.[23] It neglected the scope of digital or algorithmic competencies, which are highly needed in today’s world.[24] Although ESSA did allow for federal funding to support effective instructional technology,[25] it does not truly look into how states’ discretion could result in economic and social problems for marginalized groups. This creates variation in terms of the availability of AI-related curriculum and teacher training. The Individuals with Disabilities Education Act (IDEA) also entails AI literacy as it requires schools to provide students with disabilities with the proper technological tools to obtain a public education.[26]
AI tools are becoming deeply embedded in learning systems. IDEA’s accessibility mandates will require many school districts to evaluate algorithmic bias and ensure that these tools are meeting the federal disability standards under Section 504 of the Rehabilitation Act of 1973, which prohibits discrimination based on disability in any program or activity receiving federal financial assistance, and requires schools to provide students with equal access to educational opportunities.[27] It establishes that students with disabilities must receive the accommodations, support, and services necessary to participate meaningfully in the learning environment.[28] But there is no federal policy that explains or even details how schools should evaluate AI tools for compliance.
B. Civil Rights Protections and Anti-Discrimination Law
AI literacy policy is built on the core foundational elements of the civil rights statutes. Title VI of the Civil Rights Act of 1964 prohibits discrimination in programs that are funded by the federal government.[29] When it comes to equal access, AI-relevant coursework, broadband connectivity, or algorithmic tools could contribute to discrimination if they are disproportionately harming racial groups.[30] This would also apply to Title IX for gender discrimination[31] and the Equal Educational Opportunities Act (EEOA) for language minorities.[32]
Historically, courts have understood and recognized education inequity as a civil rights concern and problem, even when they don’t set standards for what schools would have to do to comply. The Supreme Court’s decision in Milliken v. Bradley (1974) held that schools could not be required to cross district lines to remedy segregation unless it was proven that multiple districts had deliberately engaged in discriminatory practices. In other words, the Court rejected a metropolitan-wide desegregation plan and ruled that predominantly white suburban districts did not have to participate in remedies for a segregated urban district. This ruling sharply limited the ability to address educational inequity regionally. It effectively cemented district boundaries as barriers, preventing resource-sharing or collaborative remedies across districts even when inequality was the product of broader housing patterns, regional funding structures, and metropolitan-level decisions.[33]
C. Federal and State Digital Learning Policies
Federal digital learning policy is established through the E-Rate Program, part of the Telecommunications Act of 1996, which supports school districts with broadband and telecommunications services.[34] Through the implementation of E-Rate, many schools have connectivity, but the program has not closed gaps in broadband speed, home internet access, quality, or digital training.[35] This is important to note as AI tools become more dependent on bandwidth. Bandwidth becomes the focal point of the issue because it determines whether schools can actually participate in digital or AI-enabled learning in the first place. When a district lacks sufficient bandwidth, it cannot run the tools, platforms, or instructional programs that these initiatives assume are baseline conditions. As a result, the conversation shifts from how to implement AI literacy to whether implementation is even possible in the first place. Bandwidth exposes the underlying structural divide: districts with strong, reliable internet can adopt new technologies with ease, while districts with weak or inconsistent connectivity are automatically excluded from the modernization initiative envisioned. In this way, bandwidth decides who gets to benefit from digital learning environments and who is left behind, revealing that the core issue is not innovation itself but unequal access to the infrastructure required to support it.
The Obama administration announced the ConnectEd Initiative, which was designed to modernize digital learning environments by expanding access to up-to-date classroom technologies, improving broadband connectivity, and encouraging schools to integrate digital tools into everyday instruction, but its impact was uneven because participation was never guaranteed.[36] Today, many states take different approaches to digital literacy and computer science mandates. Most districts understand the importance of implementing or providing resources that pertain to AI literacy, though some districts lack the critical resources needed to help facilitate competencies. There are no detailed standards or approaches for school districts in AI education, which contributes to reinforcing inequities in an increasingly digital world.
D. Procurement Rules and Technology Adoption in Schools
School districts can adopt AI-enabled tools through public procurement rules that are generally cost-effective, require vendor compliance, and are central to data privacy. Procurement practices are governed by federal statutes such as the Family Education Rights and Privacy Act (FERPA), which aims to regulate data collection through digital platforms.[37] States can also add additional requirements. California, for example, enacted the Student Online Personal Information Protection Act (SOPIPA), which aims to limit the way in which ed-tech vendors can use student data.[38] Despite that, procurement practices don’t necessarily require equity assessments, algorithmic bias evaluations, or community participation in technology selection. There are various districts that don’t have the necessary resources to negotiate equitable contracts or get tools that can critically follow civil rights standards. Due to this limitation, well-resourced school districts can adopt tools to address various concerns, whereas underfunded and under-resourced school districts are limited, thus they are forced to adopt low-quality tools or none at all, which reproduces the digital divide through procurement practices.
E. Emerging AI Governance and the Absence of Educational Standards
At the federal level, there are active measures to shape AI governance by way of executive guidance rather than comprehensive legislation. The Blueprint for an AI Bill of Rights (2022) put forth principles for safe and non-discriminatory algorithmic use, but doesn’t detail how this process is supposed to be actually mandated.[39] The National Institute of Standards and Technology (NIST) released an AI Risk Management Framework in 2023, which put forth ways in which risks can be mitigated, but it doesn’t pertain to schools specifically.[40] Many states have already started to look at ways in which they can govern AI, but the focus is generally applied to commercial use and employment; it does not address the importance of AI literacy implementation or guidance in K-12 education. If there is no governing statute that details AI literacy standards, data protections, or civil rights obligations in algorithmic education, AI implementation would continue to be fragmented by districts’ inconsistent decisions.
F. Political Forces Shaping the Landscape
AI literacy is not only shaped by law but also guided by the rhetoric of competing political pressures. Many technology companies understand the importance of schools as a market, so they allow schools early access to their tools. State governments oversee innovation and workforce preparation, yet this often privileges economic competitiveness over structural equity. Various civil rights organizations have already raised concerns about algorithmic discrimination, digital exclusion, and unequal access to AI. This inequality must be acknowledged as policymakers seek either to define AI literacy as a public good, requiring a universal framework to centralize the resource, or allow it to develop as a market-driven skill that benefits only communities with sufficient infrastructure. Because there is no federal approach, political and economic pressures will continue to shape AI implementation.
III. Diagnosing Structural Barriers to Equitable AI Literacy
AI policy in the United States is being built on top of an educational system that is legally and socially unequal; the foundational elements of this system ultimately shape who benefits from technological innovation. Most political decisions, generally from the federal and state levels, assume that school districts have the necessary infrastructure, staffing, and curricular flexibility to adopt AI tools, but decades of policy failures have shown that equal access has never existed. Because of these political decisions, the U.S. faces a critical risk, as marginalized students would be excluded from access, and policy decisions would prioritize students in predominantly white districts. One problem stems from resource-dependent educational federalism, which is a structure that is reinforced by the Supreme Court. In San Antonio Independent School District v. Rodriguez, the Court ruled that education was not a fundamental right under the Constitution, which allowed states to fund schools through localized property tax systems.[41] This paved the way for wealthier school districts to establish stronger educational infrastructures. This consists of STEM opportunities and broadband access, while marginalized communities face restraints. AI literacy is deeply intertwined with the long-standing battle for educational equality, a policy that neglects the historical domains that infringe upon one’s rights could ultimately reproduce the inequities that the court caused in Rodriguez.
Additionally, underserved communities are disconnected from the policies that are guiding digital learning. Although the Telecommunications Act of 1996 aimed to expand internet access, it failed to address some critical problems, such as broadband quality, device access, and technology training.[42] AI literacy deeply depends on digital environments, but if there is no detailed approach to access, under-resourced communities will suffer, and the political effects of the act will linger in the new digital domain.
Further, higher education institutions are also reproducing historical inequities. Colleges with higher endowments can address the critical concerns that pertain to AI literacy as they present themselves, while institutions with lower endowments face critical challenges to even establishing guidelines for their students. Shoval’s work is a key contribution to understanding the racialized political economy of emerging AI literacy. She argues that when institutions of higher education attempt to “prepare” students for an AI-driven future, they inadvertently reproduce the same racial and economic hierarchies embedded in technological systems. Thus, rather than closing the gap, AI literacy programs risk expanding it, aligning universities with broader capitalist logics that privilege those already advantaged.
Lastly, AI literacy frameworks are structurally blind. Many models assume that students already have the technical understanding of AI, as they focus on interpretation, data reasoning, and computational thinking.[43] Although technical understanding is highly important, they disregard the fact that some students may not have the same understanding or exposure to technological literacy. Research on the digital divide shows that AI-powered tools deepen inequality when introduced into underfunded schools.[44] The problem is very visible: AI literacy policy, which is currently being constructed, is not a neutral process. It is consistently shaped by legal and political decisions that have historically marginalized communities of educational opportunity and digital access. Policymakers must focus on the structural impact first; if not, AI literacy standards will continue to harm marginalized communities.
IV. Policy and Legal Reforms for Equity-Centered AI Literacy
In order to properly design an AI literacy framework that accounts for equity, policymakers must address the critical structural inequities that have been formulated through historical and political landscapes. AI literacy must not be centered or governed based on technical competency; it should recognize civil rights, which have been deeply embedded in constitutional, statutory, and administrative obligations. On that note, this section proposes various ways in which we could reform AI literacy policy toward racial, economic, and geographic equity. These reforms are based on research comparing different places, federal civil rights rules, and studies showing how unequal schools and technology access are across different communities.
A. Federal Mandate for Equity-Centered AI Literacy Standards
Congress should create a federal equity mandate requiring states to showcase how AI literacy implementation plans address historic inequities within underserved communities. Looking at federal obligations under the Every Student Succeeds Act (ESSA), this mandate would require states to consider how well AI literacy initiatives are also measuring equity benchmarks rather than solely focusing on innovation goals. It would ensure that AI literacy standards do not contribute to resource disparities, which are often tied to property tax-funding systems that contribute significantly to affluent school districts.
AI literacy research has shown that students who come from a background with higher parental education and socioeconomic status tend to have higher AI competencies.[45] If we don’t set federal requirements to stop these disparities from deepening, leaving this problem solely to the states would cause some school districts to be privileged over others that have the required infrastructure. Federal policy implementation thus becomes important to avoid unequal adoption patterns that have been on the rise since digital learning technologies started to become the standard in the late twentieth century.
B. Civil Rights Enforcement for Algorithmic and Educational Access
The Department of Education’s Office for Civil Rights should clarify that racially inequitable access to AI literacy may violate Title VI of the Civil Rights Act when students of color are denied equal access to educational resources, technology, and related learning opportunities. Various courts have understood this issue and recognized that resource allocation was tied to the state, which has contributed to civil rights concerns in education. Due to the fact that unequal access to AI tools has the potential to widen achievement gaps, OCR should make sure that AI literacy education is guaranteed to all students, not just the select few who have the resources to obtain literacy. OCR should also require that all school districts publicly disclose and report disparities in access to technological resources. There are already mechanisms in place to address various problems under ESSA, but it is also important to include digital and algorithmic access. Directly reporting inequitable access would force many policymakers to be accountable for disparities.
C. Federal Digital Infrastructure Funding Targeted to Underserved Districts
Federal broadband initiatives do not eliminate the other barriers that are deeply embedded in society. For example, as of 2023, the FCC reports that broadband availability in the U.S. has reached over 90%, yet nearly 28 million households still lack a home internet subscription, primarily due to affordability, device shortages, and limited digital and AI literacy rather than physical access to broadband infrastructure.[46] Moreover, the National Center for Education Statistics finds that students in high-poverty districts are 2-3 times more likely to lack access to a computer at home, even when broadband service is technically available.[47] In order to properly address this gap, it is important for Congress to put forth a funding program that allocates AI-literacy infrastructure grants to districts with high rates of poverty. This funding would address various issues, such as funding for devices, computational learning tools, and culturally responsive AI curricula. Equity is something that can be addressed when resources are provided to communities that have been historically underserved and excluded from technological innovation. If these issues are not addressed, we will accelerate and deepen the digital divide.
D. Procurement Standards That Prevent “Inequity by Design”
State and district procurement rules need to require algorithmic transparency, data protection, and equity audits for all AI-enabled educational tools. Procurement policy has historically led to educational innovations in classrooms, but it doesn’t account for the various demographics it impacts. Therefore, I suggest that there be a way to actually account for exploring ways in which standards could be created to prevent inequity by design.
E. Community-Centered Co-Design and Governance
Various studies have also alluded to the fact that AI literacy programs are most effective when they include cultural capital and the lived experiences of underserved communities.[48] In order to properly address this, states should focus on establishing community AI literacy councils; the council could be composed of parents, teachers, students, technologists, and local advocates. These councils would allow for the various people across communities to take part in the creation of standards, evaluation of curricular tools, and oversight of equity outcomes.
F. Cross-Disciplinary Integration Beyond STEM
Lastly, it is important to mandate AI literacy across the curriculum instead of solely focusing on it as STEM instruction. Various models, from history and civics to social work and the arts, allow students to break down AI’s impact in society, which can allow them to understand the social, ethical, and political implications of AI.[49] This allows AI to be regarded as a social and legal process instead of relying solely on technical understanding. By including the humanities, we allow for the structural inequities that are documented in K-12 systems to be seen and allow people to address them as they arise.[50] These proposals allow one to build and categorize what it means to address some of the political and legal issues that seem to allow inequities to continue through the development of AI, through legal, civil rights, and governance gaps. By embedding equity into federal mandates, civil rights enforcement, procurement rules, and community governance structures, policymakers can construct an AI literacy framework that expands, rather than restricts, educational and democratic opportunity.
Conclusion: AI Literacy as a Civil Rights Imperative
Artificial intelligence is a growing field in the United States; therefore, it is positioned with the effects of educational segregation, unequal funding, and uneven technological development. Various policy agendas will be put into place to increase AI literacy, but it is important that they consider the legal and political dynamics that have contributed to educational access. The existing framework within the United States is shaped by the court, federal statutes, and digital learning policies that don’t distribute technological infrastructure equally across the nation. These are the conditions in which AI literacy is going to be built from, so these domains are to be taken into account by any policymaker who attempts to implement good policy to enrich and better the lives of others.[51] If policymakers neglect history, they will be contributing to a new formalization of disparities that stems from the rhetoric of innovation, which is usually positioned as solving problems.
AI is not only a technical tool; it has been deeply embedded into our social lives as it contributes to access to employment, information, civic participation, and economic mobility. If we implement policies that neglect how technological innovation only allows communities that are advantaged to take part, we will be infringing upon minorities’ informational, economic, and political rights. The inequities that were identified throughout this piece, such as unequal access to AI adoption, further allude to the idea that AI literacy is being built within a system that has already contributed to asymmetrical opportunity. The proposals that were offered within this article would allow us to move from an act of simple policy commitments to real action that is built on the basis of inclusion.
The legal and policy proposals offered in this Article are designed to move beyond rhetorical commitments to inclusion. This would require us to reimagine the actions of federal and state authorities, pushing for policy implementation that aims to redesign procurement and funding structures and build new frameworks that are central to making sure the distribution of resources is equitable. This would require community experts to be at the forefront of issues that are particularly disadvantageous to marginalized groups, as they would also be included in the design, evaluation, and governance of AI-related programs.
If policy implementation does not seek to incorporate the voices of the marginalized into the practical aspects of AI implementation, then AI literacy will only benefit the people who already have access to it, leaving inequality unaddressed as the central issue in the American political context. Ultimately, the future is in the hands of policymakers. If they prioritize innovation and competition over democratic empowerment, society will eventually ask how such inequities could have been prevented. With policies built on equity, participatory governance, and historically informed policymaking, AI literacy could contribute to the rise of America’s global infrastructure, as marginalized students would be able to contribute to America’s technological innovation and help shape AI leadership.
[1] National Center for Education Statistics, Forum Guide to Digital Equity (2024).
[2] Bruce Mitchell & Juan Franco, HOLC “Redlining” Maps: The Persistent Structure of Segregation and Economic Inequality 4 (2018), https://ncrc.org/wp-content/uploads/dlm_uploads/2018/02/NCRC-Research-HOLC-10.pdf.
[3] See Richard Rothstein, The Color of Law: A Forgotten History of How Our Government Segregated America 92–130 (2017).
[4] See Christopher Silver, The Racial Origins of Zoning in American Cities, in Urban Planning and the African-American Community: In the Shadows .
[5] Milliken v. Bradley, 418 U.S. 717 (1974).
[6] Id.
[7] Federal Communications Commission, E-Rate and Education (A History), https://www.fcc.gov/general/e-rate-and-education-history.
[8] Id.
[9] The White House, Educate to Innovate, https://obamawhitehouse.archives.gov/issues/education/k-12/educate-innovate; President Obama Launches “Educate to Innovate” Campaign for Excellence in Science, Technology, Engineering & Math (Stem) Education, Nov. 23, 2009, https://obamawhitehouse.archives.gov/the-press-office/president-obama-launches-educate-innovate-campaign-excellence-science-technology-en; U.S. Department of Education, Fundamental Change: Innovation in America’s Schools Under Race to the Top 2, 6 (2015), https://files.eric.ed.gov/fulltext/ED577009.pdf.
[10] Race to the Top Fund, 74 Fed. Reg. 59,836, 59,841–42, 59,844–45 (Nov. 18, 2009) (describing selection criteria, including state capacity, data systems, and effective teachers and principals); Legis. Analyst’s Off. 7, Overview of Federal Race to the Top Grants (Sept. 29, 2009), https://lao.ca.gov/handouts/education/2009/Overview_of_Federal_Race_to_the_Top_Grants_92909.pdf (listing “data systems that support instruction” and “effective teachers and principals” among reform areas and noting capacity-based criteria); Fundamental Change, supra note 9, at 2, 6; Sylvia A. Allegretto, Public Education Funding in the U.S. Needs an Overhaul, Econ. Pol’y Inst. (July 12, 2022), https://www.epi.org/publication/public-education-funding-in-the-us-needs-an-overhaul/ (explaining that school funding relies heavily on state and local resources, particularly property taxes, and remains inequitable); An Analysis of School Funding Equity Across the U.S. and Within Each State, Educ. Tr. 1–2 (Dec. 2022), https://edtrust.org/wp-content/uploads/2014/09/Equal-Is-Not-Good-Enough-December-2022.pdf (finding that districts serving more students from low-income backgrounds and students of color often receive less state and local revenue); Daniel A. Lukes & Dora C. Costa, The Lingering Legacy of Redlining on School Funding, School Segregation, and Gentrification in the U.S., https://files.eric.ed.gov/fulltext/ED616673.pdf (linking historical redlining to present-day disparities in school and district funding).
[11] Yurong Guan et al., AI Literacy Predicts Computational Thinking through Multidimensional Interactions among Chinese High School Students, 15 Sci. Rep. (2025).
[12] Dorit Hadar Shoval, Artificial Intelligence in Higher Education: Bridging or Widening the Gap for Diverse Student Populations?, 15 Educ. Sci. 637 (2025).
[13] U.S. Department of Education, No Child Left Behind: A Desktop Reference (2002), https://www.ed.gov/media/document/no-child-left-behind-desktop-reference-archived-2002-98652.pdf.
[14] Every Student Succeeds Act, Pub. L. No. 114-95, §§ 4108(3), 4109(a), 129 Stat. 1802, 1958–59 (2015) (hereinafter ESSA); See U.S. Department of Education, Overview of Final Regulations: Accountability, Data Reporting, and State Plans, (2016), https://www.ed.gov/media/document/accountstateplan12816pdf-55349.pdf; U.S. Department of Education, Non-Regulatory Guidance: Student Support and Academic Enrichment Grants (2016), https://www.ed.gov/sites/ed/files/2020/07/essassaegrantguid10212016.pdf.
[15] See, e.g., Allegretto, supra note 10; Megan Kuhfeld et al., Understanding the Initial Educational Impacts of COVID-19 on Communities of Color (2021), https://files.eric.ed.gov/fulltext/ED617855.pdf; Amanda Bergson-Shilcock & Roderick Taylor, Closing the Digital Skill Divide: The Payoff for Workers, Business, and the Economy (2023), https://files.eric.ed.gov/fulltext/ED626621.pdf.
[16] Gopal K. Singh et al., Digital Divide: Marked Disparities in Computer and Broadband Internet Use and Associated Health Inequalities in the United States, 4 Int. J. Transl. Med. Res. Public Health (2020).
[17] Id.
[18] Emily A. Vogels, Digital Divide Persists Even as Americans with Lower Incomes Make Gains in Tech Adoption, Pew Research Center, Jun. 22, 2021, https://www.pewresearch.org/short-reads/2021/06/22/digital-divide-persists-even-as-americans-with-lower-incomes-make-gains-in-tech-adoption/.
[19] Id.
[20] See generally Organisation for Economic Co-operation and Development, A Broken Social Elevator? How to Promote Social Mobility (2018), https://www.oecd.org/content/dam/oecd/en/publications/reports/2018/06/a-broken-social-elevator-how-to-promote-social-mobility_g1g8e196/9789264301085-en.pdf; Hannah Borhan, Civic Education as a Pathway to Inclusive Societies: Exploring the Role of Education in Fostering Civic and Social Engagement, (2025), https://www.oecd.org/content/dam/oecd/en/publications/reports/2025/02/civic-education-as-a-pathway-to-inclusive-societies_6d52238a/3f128be8-en.pdf; Organisation for Economic Co-operation and Development, AI and Work, https://www.oecd.org/en/topics/ai-and-work.html; Organisation for Economic Co-operation and Development, AI and the Future of Social Protection in OECD Countries, (2025), https://www.oecd.org/content/dam/oecd/en/publications/reports/2025/06/ai-and-the-future-of-social-protection-in-oecd-countries_038f49ed/7b245f7e-en.pdf; Organisation for Economic Co-operation and Development, Governing with Artificial Intelligence: The State of Play and Way Forward in Core Government Functions (2025), https://doi.org/10.1787/795de142-en.
[21] See generally Bill Pitkin, Katharine Elder & Danielle DeRuiter-Williams, Building a Housing Justice Framework (2022), https://www.urban.org/sites/default/files/2022-08/Building%20a%20Housing%20Justice%20Framework.pdf; Boshart, How Tenant Screening Services Disproportionately Exclude Renters of Color from Housing, Housing Matters, Dec. 21, 2022, https://housingmatters.urban.org/articles/how-tenant-screening-services-disproportionately-exclude-renters-color-housing; Brendan Chen et al., Oversight of Tenant Screening Companies May Change. Here’s What Landlords and Renters Should Know to Protect Themselves, Housing Matters, Apr. 9, 2025, https://www.howhousingmatters.org/articles/oversight-tenant-screening-companies-may-change-heres-what-landlords-and-renters-should; Judah Axelrod et al., Opening the “Black Box” of Tenant Screening: Analyzing Data Matches in Court Data (2025), https://www.urban.org/sites/default/files/2025-03/Opening-the-Black-Box-of-Tenant-Screening.pdf.
[22] Dhruv Mehrotra et al., How We Determined Crime Prediction Software Disproportionately Targeted Low-Income, Black, and Latino Neighborhoods, The Markup, Dec. 2, 2021, https://themarkup.org/show-your-work/2021/12/02/how-we-determined-crime-prediction-software-disproportionately-targeted-low-income-black-and-latino-neighborhoods.
[23] Elementary and Secondary Education Act of 1965, Pub. L. No. 89-10, 79 Stat. 27 (1965), amended by Pub. L. No. 114-95, 129 Stat. 1802 (2015), https://www.ed.gov/sites/ed/files/documents/essa-act-of-1965.pdf; Overview of Final Regulations, supra note 14.
[24] See generally ESSA.
[25] Id. § 4101.
[26] Individuals with Disabilities Education Act, 20 U.S.C. §§ 1400–1482.
[27] 29 U.S.C. § 794 (2024); U.S. Department of Education, Section 504, https://www.ed.gov/laws-and-policy/individuals-disabilities/section-504 (last visited Mar. 16, 2026); U.S. Department of Education, Disability Discrimination: Equal Access to Opportunities, https://www.ed.gov/laws-and-policy/civil-rights-laws/disability-discrimination/disability-discrimination-key-issues/disability-discrimination-equal-access-opportunities (last visited Mar. 16, 2026); Miguel A. Cardona, Roberto J. Rodríguez & Kristina Ishmael, Artificial Intelligence and the Future of Teaching and Learning: Insights and Recommendations (2023), https://www.ed.gov/sites/ed/files/documents/ai-report/ai-report.pdf; John Ross, Guidance for Developing Policies to Govern the Adoption and Use of Artificial Intelligence in K-12 Schools, https://files.eric.ed.gov/fulltext/ED655341.pdf.
[28] Rehabilitation Act of 1973 § 504, 29 U.S.C. § 794.
[29] Civil Rights Act of 1964, tit. VI, 42 U.S.C. § 2000d.
[30] Alexander v. Sandoval, 532 U.S. 275 (2001) (limiting private disparate-impact suits but affirming agency enforcement authority).
[31] Education Amendments of 1972, tit. IX, 20 U.S.C. §§ 1681–1688.
[32] Equal Educational Opportunities Act of 1974, 20 U.S.C. §§ 1701–1758.
[33] Milliken v. Bradley, 418 U.S. 717 (1974).
[34] Telecommunications Act of 1996, Pub. L. No. 104-104, § 254(h), 110 Stat. 56, 91-92 (1996) (hereinafter Telecommunications Act).
[35] Federal Communications Commission, supra note 7; Federal Communications Commission, E-Rate – Schools & Libraries USF Program, https://www.fcc.gov/general/e-rate-schools-libraries-usf-program (last visited Mar. 18, 2026); Modernizing the E-Rate Program for Schools and Libraries, 29 FCC Rcd. 8870, 8871–72 (2014) (noting the need for greater high-speed broadband capacity in schools and libraries); Permitting E-Rate Program Support for Off-Premises Use of Wi-Fi Hotspots and Services, FCC 24-76, ¶¶ 1–5 (July 29, 2024), https://docs.fcc.gov/public/attachments/FCC-24-76A1.pdf (recognizing that many students still lacked off-premises internet access and remained on the wrong side of the digital divide); Nicol Turner Lee, Bridging Digital Divides between Schools and Communities (2020), https://www.brookings.edu/articles/bridging-digital-divides-between-schools-and-communities/.
[36] The White House, ConnectED, https://obamawhitehouse.archives.gov/issues/education/k-12/connected (last visited Mar. 18, 2026).
[37] Family Educational Rights and Privacy Act, 20 U.S.C. § 1232g.
[38] Student Online Personal Information Protection Act, Cal. Bus. & Prof. Code §§ 22584–22585.
[39] See generally White House Office of Science and Technology Policy, Blueprint for an AI Bill of Rights: Making Automated Systems Work for the American People (2022), https://www.govinfo.gov/content/pkg/GOVPUB-PREX23-PURL-gpo193638/pdf/GOVPUB-PREX23-PURL-gpo193638.pdf.
[40] See generally National Institute of Standards and Technology, Artificial Intelligence Risk Management Framework (AI RMF 1.0) (2023), https://doi.org/10.6028/NIST.AI.100-1.
[41] San Antonio Indep. Sch. Dist. v. Rodriguez, 411 U.S. 1 (1973).
[42] Telecommunications Act.
[43] Davy Tsz Kit Ng et al., Conceptualizing AI Literacy: An Exploratory Review, 2 Comput. Educ. Artif. Intell. (2021).
[44] Lourdu Vesna et al., Digital Divide in AI-Powered Education: Challenges and Solutions for Equitable Learning, 10 J. Inf. Syst. Eng. Manag. 300 (2025).
[45] Guan et al., supra note 11.
[46] National Telecommunications and Information Administration, Federal Broadband Funding Report: Investing in Internet for All (2023), https://www.ntia.gov/sites/default/files/reports/federal_broadband_funding/2023-federal-broadband-funding-report.pdf.
[47] See Institute of Education Sciences, Use of Educational Technology for Instruction in Public Schools: 2019—20 (2021), https://nces.ed.gov/pubs2021/2021017.pdf.
[48] See, e.g., Huajie Jay Cao et al., AI Literacy for Underserved Students: Leveraging Cultural Capital from Underserved Communities for AI Education Research, CHI Conf. Hum. Factors Comput. Syst. (2025).
[49] See, e.g., Eunhye Ahn et al., Artificial Intelligence (AI) Literacy for Social Work: Implications for Core Competencies, 16 J. Soc. Soc. Work Res. 9 (2025).
[50] See generally Katarina Sperling et al., Perspectives on AI Literacy in Middle School Classrooms: An Integrative Review, Postdigital Sci. Educ. (2025).
[51] Brown v. Board of Education, 347 U.S. 483 (1954).