As a new school year begins, some classrooms may use artificial intelligence tools to flag a student struggling with classwork, track attendance, or try to raise reading scores.
But as school districts make decisions on which tools are safe and worth spending thousands or millions of dollars on, some officials feel overwhelmed by the array of choices. Some states and districts have begun offering guidance for vetting AI purchases, but educators and experts say districts are still carrying much of the burden on their own.
While school districts have long wrestled with education technology purchases, AI is different because the industry is moving faster than districts and states can keep up with, said Mark Schneider, a nonresident senior fellow at the conservative-leaning think tank the American Enterprise Institute.
“There’s always an asymmetry between what the providers know and what the districts know,” Schneider told Stateline. “And the question is, can we set up mechanisms for helping the districts understand what’s going on and choosing better.”
Billions of dollars are at stake. The overall U.S. education technology market—which includes computers, software, and digital content— generated almost $48 billion in revenue in 2024 and is expected to reach more than $90 billion by 2030, according to Grand View Research. The AI in education market in the United States generated about $2.5 billion in revenue last year and is expected to exceed $15 billion by 2033, the group said.
Scott Langford, superintendent of the Sumner County schools in Tennessee, said his district decided to test AI to address plateauing middle school reading scores. Langford said the district wanted to know whether AI could give students immediate feedback and help teachers see, in real time, where students struggled during regular classroom instruction.
“You have to identify what you need,” Langford said. “The (AI ed-tech) market is so overwhelming for districts and for vendors. … Most of them don’t really do much that drives any kind of change for students.”
Sumner first tried Coursemojo—an AI-powered literacy tool used to support classroom instruction and provide real-time feedback to students and teachers—in a five-day, micro-pilot in a 6th grade classroom, then piloted the tool in half of its 6th grade classrooms. Langford said the strongest tools are not the ones that replace teachers, but the ones that “help them see student understanding more quickly.”
Sumner has continued using Coursemojo after the pilot, Langford said.
The Allentown School District, a large, urban district in Pennsylvania, uses a two-phase review process for AI programs that Superintendent Carol Birks said includes non-negotiables such as student safety and data security. Vendors must ensure that all user-generated data and chat logs remain the exclusive property of the district, and vendors must explicitly state that student data will not be sold or used to train AI models.
Birks said that while districts are making these decisions on AI tools, state and federal guidance would be helpful. In addition to Coursemojo, the district also uses other AI platforms like Google’s Gemini.
“A significant challenge is that the onus of making these decisions is placed entirely back on the individual school district,” said Birks. “Although Allentown has built strong internal capacity to handle these responsibilities, guidance from the state level on artificial intelligence vetting or protections would greatly assist school districts.”
What vetting looks like
In most cases, vendors must sign a data processing agreement, complete a security questionnaire and pass legal, privacy, and security reviews before a tool can be approved for use in schools. And some states already have laws governing traditional education technology purchases and student data protections. But many districts haven’t laid out best practices or recommendations for how school districts enter contracts with vendors or even bidding processes for certain tools.
And some AI use in classrooms has drawn criticism and legal action. Students and parents in Arizona and Kansas have filed federal lawsuits challenging school districts’ use of Gaggle—a student-safety monitoring platform that scans activity on school-issued accounts and devices—saying it violated their constitutional rights. In Los Angeles, the superintendent of public schools resigned after the district paid $3 million for an AI chatbot from a company called AllHere, which later filed for bankruptcy and whose founder was charged with fraud.
State education boards and agencies, which already regulate contracts for textbooks, transportation, and student data systems have filled in some gaps with vetting and procurement guidance for districts.
The Pennsylvania Department of Education instructs schools to understand who controls the data entered into any AI tool, examine third-party data-sharing practices, limit data collection to what is necessary, keep a human “in the loop” if the tool is being used for grading and discipline, and assess whether it improves student performance.
New York City Public Schools updated its privacy and security review process to include standards for AI, require vendors to disclose what their AI tools are capable of, prohibit the use of student data to train AI models, and meet transparency requirements so tools can be explained to families and students.
Chicago Public Schools released an AI Guidebook governing the use of AI software and models and in December began blocking unapproved third-party AI products from its network.
“Chicago Public Schools welcomes the promising innovation of artificial intelligence in teaching and learning, while safeguarding the development of students’ foundational skills and critical thinking,” a Chicago Public Schools spokesperson wrote to Stateline.
The U.S. Department of Education has provided little guidance to districts about AI purchases.
The Southern Regional Education Board—a consortium of 16 Southern states that provides guidance on school issues—published an AI procurement and evaluation checklist last year for K-12 schools.
Leslie Eaves, program director of project-based learning at the board who co-led the AI commission, said districts should first let adults test AI tools before placing them in front of students. She cautions that teachers should review AI-generated material before using it in the classroom and they need to trust their experience and expertise.
“Anything with AI is a risky click,” she said. “There’s a lot of tool makers out there that are going to sell you their product and they’re going to say it’s the biggest and best in everything … At the end of the day, humans are the ones who will know what’s best for their schools and students.”’
Janice Mak, an associate director of curriculum and pedagogy at Arizona State University’s Learning Engineering Institute, says districts should evaluate whether an ed-tech tool is accessible, explainable, grounded in learning science, and solving a real educational need.
“State level policy or guidance only goes so far, and there’s also a research gap and there’s also a policy gap,” Mak told Stateline.
Mak said she heard from a middle school teacher piloting an AI tool who noticed English learners were being unfairly penalized or misread by the system. The students noticed the bias, she said, and the teacher used it to teach them about algorithmic bias and how AI systems are trained.
‘A tall order for districts’
Still, districts are being asked to do too much on their own, said Sue Thotz, director of outreach for Common Sense Education at Common Sense Media, a nonprofit that researches and evaluates media and technology used by children and schools.
Thotz points to a California data privacy agreement, which districts can use to ensure vendors comply with the state’s student data privacy laws, as a way to take some burden off districts by creating a blanket agreement vendors sign onto. The California Student Privacy Alliance, which developed the agreement, gives school systems a searchable database of vendors that have signed, reducing the need for each district to negotiate privacy terms from scratch.
“Placing the burden upon a district to make sure that every single vendor they’re working with is in full compliance with their state and federal laws when it comes to privacy, that’s a tall order for districts that are already stretched very, very thin,” she said.
Districts may also be sorting an accumulation of a decade’s worth of technology purchases, said Tracy Weeks, senior director of education policy and strategy at Instructure, the education technology company that makes the Canvas learning management system.
An Instructure report found districts had access to an average of 3,001 digital tools, while students and educators used only four of those tools on average. The report, based on Canvas launch data from more than 12.6 million K-12 users, said the volume of available tools has teachers navigating too many platforms and districts overseeing student data across vendors.
“What I think has happened is that I think we did a lot of digital hoarding,” Weeks said. Now, she said, districts are asking, “What are the things we should actually keep? What do we need to use? What can actually move the needle?”
Shrinking budgets and the expiration of federal pandemic-era funding has also forced districts to scrutinize purchases.
Dacia Toll, co-founder of Coursemojo, the software used by Sumner County and Allentown, said districts should first ask whether an AI product aligns with what is known about cognitive science and strong teaching methods.
“Education has had a long-standing issue with vetting ed tech for safety, usability, and efficacy, and I just think AI is making that more pronounced,” she told Stateline. “There are a handful of additional risks, and there’s so much hype right now that I think some districts are feeling like they’re behind if they’re not actively implementing AI.”
