As U.S. schools explore ways to integrate emerging tech into daily operations, a growing number of special education teachers are using generative AI to help write and manage documentation for students with disabilities, according to a recent report by the nonprofit Center for Democracy and Technology (CDT).
The findings point to a fast-moving shift — one that brings both opportunities and risks — for educators who draft Individualized Education Plans (IEPs), or legal documents that outline how an educational program for a child with disabilities is designed to meet their specific needs.
While many teachers say AI helps reduce work time, CDT researchers warn that its use in this legally mandated process could compromise student privacy, reinforce bias and weaken the personalized nature of supports for students with disabilities required under federal law.
“The increased use of AI in this context is not surprising, considering the techno-optimist rhetoric surrounding the use of AI in schools, its potential usefulness, and special education teachers’ extreme workload, which often leads to burnout,” the report said. “But it is not clear that teachers, administrators and students are aware of the potential risks of using AI in this context … without negatively impacting disabled students and potentially even violating their legal rights.”
WHAT IS AN IEP?
Under the federal Individuals with Disabilities Education Act (IDEA), any public school student in grades K-12 who has a qualifying disability and receives special education or related services is required to have an IEP. The document outlines the specific supports, services and accommodations the student needs in order to receive an inclusive and effective education each year.
Developing and implementing an IEP is intended to be a personalized and highly collaborative effort between guardians, teachers, related service providers, and at times, the student, the report said. The legally binding document also includes highly sensitive information about students’ strengths, disabilities, academic histories and health.
IEPs are more than checklists, though, said Ariana Aboulafia, CDT’s project lead for disability rights in technology policy, who authored the report.
“IEPs are not only deeply personal,” she said. “They are extremely important.”
Data from the 2022-23 school year shows 7.5 million students received services under IDEA, according to the report, indicating that nearly 15 percent of all U.S. public school students rely on IEPs to access the general education curriculum in schools.
WHAT AI IS BRINGING TO IEP DEVELOPMENT
Educators are increasingly using generative AI tools to support IEP development, Aboulafia emphasized in the report, noting that “more than half of licensed special education teachers report that they use AI in some way to help them develop IEPs.”
Of the educators polled by CDT, 15 percent said they use AI to write an IEP or 504 plan in full — nearly double the number from the previous year.
“To be clear, those numbers do not mean that that is the percentage of teachers who are using AI to develop IEPs in their totality,” Aboulafia added. “They might be using it just for the narrative or just for some other portion of it, to brainstorm ideas for accommodations or that sort of thing.”
According to CDT’s findings, 31 percent of special education teachers use AI to identify trends in student progress and help determine patterns for goal setting. Thirty percent said they use it to summarize the content of an IEP or 504 plan, 28 percent to help choose specific accommodations, and 21 percent to build out the narrative portion.
Aboulafia said time-saving benefits appear to be a primary motivation, adding that teachers who use AI tools weekly “may save … up to six weeks over the course of any given school year.” According to the report, AI can generate templates or drafts with minimal prompting, reducing the need to build each IEP from scratch.
Amid persistent teacher shortages and high special education caseloads, the CDT said teachers may turn to AI tools to alleviate workload pressures.
The report also suggested that AI could help improve communication by refining IEP language to make it “strengths-oriented, accessible, understandable and clear for parents, students and other stakeholders,” which may reduce miscommunication and promote greater parent engagement.
As well, there is a belief that AI could aid with legal compliance and help ensure IEPs are sufficiently individualized under IDEA. If trained and properly tested for bias, such tools might recommend modifications or accommodations based on the child’s profile, according to the report.
CORE RISKS AND CONCERNS
Despite its potential advantages, Aboulafia said that AI-assisted IEP development raises serious concerns about individualization, accuracy, privacy and overreliance.
One major risk, according to the report, involves legal compliance with IDEA’s requirement for individualized instruction and measurable goals: “Certain AI tools develop IEPs that can be copied and pasted in full, and do so based on very little student-specific information. Using these sorts of IEPs, especially without significant editing, likely would not satisfy the individualized requirements of IDEA.”
Accuracy and bias constitute another concern. The report warned that biases in training data could reinforce stereotypes or inequitable recommendations, especially since students with disabilities are often underrepresented in data sets used to train large language models.
“These AI tools show racial bias. They show all sorts of other biases,” Aboulafia said. “There’s all sorts of things written about the algorithmic bias in these sorts of tools, meaning that the bias could be coded into the IEP.”
Aboulafia added that privacy and data protection are at stake.
“Under FERPA, some information can be disclosed to contractors,” she said, “which is why it matters whether or not the school has an agreement with an AI vendor as opposed to, let’s say, a teacher using a consumer-facing general purpose GenAI tool like ChatGPT that you would just log into on your computer.”
IMPLEMENTATION AND OVERSIGHT RECOMMENDATIONS
To address these risks, CDT stressed that human oversight is paramount when using AI in IEP development, so schools should adopt strong governance, transparency and accountability mechanisms.
Aboulafia recommended that districts consult their compliance offices and attorneys regarding potential legal risks of using AI and IEP development.
“Call your lawyers,” she said.
Aboulafia also said that even if teachers and administrators understand AI tools, their limitations and prompt-engineering basics, district policies should also include guidance on legal compliance with IDEA, Family Educational Rights and Privacy Act (FERPA) and state privacy laws.
Transparency with families is another key recommendation outlined in the report: “Whenever AI is used at any point in IEP development, we recommend that that be communicated proactively, in writing, ideally, and ideally in advance of IEP meetings, as opposed to someone showing up to a meeting or finding out afterwards,” CDT’s report said.
Aboulafia also addressed steps vendors can take on the developer side to prioritize privacy and equity practices.
“If you have the transparency and the pre-deployment audits, and the prioritization of privacy measures like data minimization or purpose limitation, and you have the active inclusion of disabled people in your development process, you may in fact just build a better product,” she said. “And so then, again, anyone who’s using [the IEP] will be able to benefit a little bit … from those more privacy-protective structures.”
