Across the United States, universities are banning AI.

Not targeted restrictions. Not nuanced policies that distinguish between using AI to write an essay and using AI to understand one. Blanket bans — institution-wide prohibitions that treat all AI use as cheating, regardless of context, regardless of need.

For most students, this is an inconvenience. For students with disabilities, it may be illegal.

The accommodation gap

Under Section 504 of the Rehabilitation Act, every college and university that receives federal funding — which is virtually all of them — must provide reasonable accommodations for students with disabilities. This includes ensuring that course materials and communication are accessible.

Under the ADA, this obligation extends to private universities as well.

For decades, this has meant things like extended test time, note-taking services, screen readers, text-to-speech software, and alternative format materials. The principle is straightforward: students with disabilities have the right to access the same educational content as their peers, using tools that bridge the gap between how information is presented and how they can process it.

AI comprehension tools are the latest version of that bridge.

A student with dyslexia who asks an AI to define jargon in a textbook passage is not cheating. She is doing exactly what a student using a dictionary or a text-to-speech tool is doing — accessing the content through a tool that makes it comprehensible to her.

A student with ADHD who asks an AI to break a 40-page reading into structured summaries with key concepts highlighted is not bypassing the assignment. He is organizing information in a way his executive function requires — the same thing a tutor or a study guide would provide.

A student with a traumatic brain injury who asks an AI to explain what a homework prompt is actually asking is not seeking an unfair advantage. She is trying to understand the question before she can answer it.

Blanket AI bans make no distinction

When a university says "no AI tools," it does not say "except for students who use them as accessibility tools." It does not route disability services offices to evaluate AI as an accommodation. It treats a student using AI to understand a reading assignment identically to a student using AI to generate a term paper.

"If a tool helps level the playing field for students with disabilities or other disadvantages, ignoring it is not neutrality — it's negligence."

— Inside Higher Ed, November 2025

The peer-reviewed "Life Raft" study, published at the ACM's Australian HCI conference in 2025, followed 13 adults with ADHD as they used ChatGPT over seven days. Participants described using AI to bridge communication gaps between neurotypical and neurodivergent thinking, break down tasks that executive function made overwhelming, and draft communications they couldn't start on their own. The researchers noted the tool was "adopted by, rather than designed for, the ADHD community" — people found it because they needed it, not because someone marketed it to them.

What happens when the tool they found and depend on is banned at their university?

The compliance risk is real

The Office for Civil Rights (OCR) at the Department of Education enforces Section 504. OCR has consistently held that:

  • Disability accommodations must be determined on an individualized, case-by-case basis
  • Blanket policies that categorically deny accommodations are presumptively non-compliant
  • The institution bears the burden of demonstrating that a particular accommodation would constitute a fundamental alteration of the program

A blanket AI ban that makes no exception for disability-related comprehension use fails the individualized-assessment requirement. A university claiming that allowing AI comprehension tools for a student with dyslexia "fundamentally alters" the academic program would face a steep burden — particularly when the student is using AI to understand the material, not to produce assessed work.

This is not a novel legal theory. It is the application of existing OCR enforcement principles to a new tool. The legal question is the same one disability services offices have been answering for decades: does this student need this tool to access the educational program on equal terms?

What's happening now

Some institutions are beginning to recognize the problem. A small but growing number of universities are:

  • Updating their AI policies to include explicit disability-accommodation exceptions
  • Routing AI-as-accommodation requests through existing disability services processes
  • Training faculty to distinguish between AI-for-generation and AI-for-comprehension

But the majority have not. Most AI policies were written rapidly in 2023-2024, in response to the initial wave of generative AI concern. They were designed to prevent academic dishonesty. Accessibility was an afterthought — or not a thought at all.

What AI Access Alliance is building

We are developing resources specifically for the education sector:

  1. Model AI accommodation policies that universities can adopt, distinguishing between AI-for-comprehension (accessibility use) and AI-for-generation (academic integrity concern).
  2. A legal brief for disability services offices explaining the Section 504 obligations that apply when a student requests AI comprehension tools as an accommodation.
  3. Student-facing resources explaining how to request AI as a disability accommodation through existing institutional processes.
  4. Faculty guidance on how to write AI policies that preserve academic integrity while respecting accommodation obligations.

The goal is not to override universities' legitimate concerns about academic integrity. The goal is to ensure that disability accommodations are not collateral damage of those concerns.

A blanket AI ban is a policy choice. It is a choice that has consequences for students with disabilities. And under existing law, institutions have an obligation to consider those consequences before making that choice.

Sources