AI in New York schools: when the promise of safety clashes with the civil rights of a million students

By Zendoric · June 28, 2026.

Nearly one million students attend New York's public schools every day. For the NAACP Legal Defense and Educational Fund (LDF) —the organization that litigated the historic Brown v. Board of Education case— that figure makes the city the country's most relevant arena for debating where surveillance technology in educational settings should be heading. On June 26, the organization submitted written testimony to the New York City Council's Education and Technology Committees, as part of a joint oversight hearing on artificial intelligence, student data and privacy in the New York City Public Schools (NYCPS).

The LDF's message was direct: halt the expansion of AI systems in public schools, including facial recognition, incident-prediction software and automated decision-making tools. The organization addressed its testimony to City Council Speaker Amanda Farías and to the chairs of both committees, Eric Dinowitz and Jennifer Gutiérrez De La Rosa, while also urging Mayor Zohran Mamdani to reject those proposals.

**The underlying problem: bias amplified by algorithms**

The LDF does not oppose technology per se, but rather its deployment in a context where inequalities are already documented. The testimony cites research showing that Black students are more than four times as likely as white students to attend schools with the highest levels of surveillance. Introducing facial recognition systems or automated decisions in that environment does not neutralize the pre-existing bias: according to the LDF's argument, it encodes and scales it.

This is not a hypothetical risk. As sector context, the most-cited studies on facial recognition —including those from the MIT Media Lab— have demonstrated significantly higher error rates for people with darker skin, especially women. When that technology is applied to high-volume settings (thousands of daily entries into school buildings), even a small error rate translates into dozens of misidentifications with real consequences for minors.

The LDF specifically notes that the New York Police Department (NYPD) —which remains responsible for security at many schools— already uses facial recognition and is piloting additional AI tools. Although the NYCPS has issued guidance prohibiting the use of AI for disciplinary decisions, the NYPD's continued presence in schools creates, according to the organization, a governance gap that is difficult to close with internal policies.

**The numbers of the trade-off**

One of the most concrete arguments in the testimony has to do with resource allocation. According to the LDF, simply halting the hiring of new school safety agents could free up approximately 90 million dollars annually. That figure, the organization argues, should be redirected toward school climate coordinators, social workers, mental health services and restorative justice programs.

It is a legitimate trade-off that deserves to be taken seriously beyond the ideological debate. The evidence on restorative justice programs in educational settings is mixed but promising: several studies in U.S. urban districts show reductions in suspensions and in the so-called 'school-to-prison pipeline' when they are implemented with sufficient resources and trained staff. The problem is that those programs are less politically visible than a security camera, and harder to evaluate in the short term.

**The First Amendment and the chilling effect**

The testimony introduces a dimension that usually falls outside the technical debate: the chilling effect on freedom of expression. In a context of intensifying immigration law enforcement and greater scrutiny of student activism, the LDF warns that expanding perimeter surveillance of schools could restrict protest and student organizing protected by the First Amendment. The organization also criticizes pending municipal legislation that would expand police powers around school buildings.

This argument connects with a broader concern in the literature on surveillance technology: the chilling effect on perfectly legal behaviors. Knowing one is being monitored changes conduct, even when nothing wrong has been done. In the case of minors in the midst of developing their civic identity, that effect can have pedagogical and democratic consequences that go far beyond immediate security.

**Where is the right place for AI in schools?**

It would be a mistake to read this debate as a blanket condemnation of artificial intelligence in education. The LDF does not say that AI has no place in classrooms: in fact, its testimony specifically mentions concern over the 'improper replacement of trained educators,' which implies that there are uses of AI that, in its view, would indeed be acceptable if implemented with adequate safeguards.

In general, the EdTech sector has spent years exploring AI applications that can improve outcomes: adaptive learning systems, early detection of reading difficulties, formative feedback tools. The critical distinction is between AI in the service of learning —student-centered, transparent, reviewable by educators— and AI in the service of control, oriented toward perimeter security or behavior prediction, with a disproportionate impact on already marginalized populations.

New York has a unique opportunity to establish a benchmark framework. With one million students and an extraordinarily rich educational data infrastructure, the city could articulate clear principles: what types of AI systems can be used in schools, under what oversight conditions, with what bias-audit mechanisms and with what rights of appeal for students and families. That framework, if built well, could become a model for other districts across the country.

**The signal that matters**

Beyond what the New York City Council decides, this episode is a clear signal that the conversation about responsible AI is no longer confined to the labs of the big tech companies or to forums specializing in algorithmic ethics. It has reached parliamentary hearings, civil rights organizations with decades of litigation behind them, and the families of students who every morning pass through metal detectors at the entrance to their schools.

The LDF's argument —that real security is built with investment in people, not in cameras— is not anti-modern. It is a call to apply the same rigor we demand of any technology: to prove that it works, for whom it works and at what cost. If an AI surveillance system cannot pass those three questions in an environment as sensitive as a public school, the burden of proof falls on those who propose it, not on those who question it.