A New Mexico education official on Wednesday pointed to the state's AI-powered reading assessment tool as one factor behind recent literacy gains, even as a legislative committee and a growing group of parents voice concerns about the technology's accuracy, privacy protections, and the lack of state guardrails. The presentation in Taos, where Lynn Vasquez, senior manager of assessment, research, evaluation and accountability for the state Public Education Department, appeared alongside Amira Learning CEO Mark Angel, followed a committee report that flagged "policy gaps" in the state's procurement of AI tools.
Vasquez said a system of early literacy assessments "leads to identification of learning needs, targeted interventions and ultimately stronger literacy outcomes," but she stressed that multiple factors were at play. "We're attributing that to a series of events, not just an assessment alone," she said. State and federal education officials have argued the technology could help reverse New Mexico's long-standing struggles with reading, though the committee's report and a parent opt-out movement are pushing back.
Parent concerns and technology struggles
Cassie Fleming, a Santa Fe parent, said her 7-year-old daughter came home "devastated" after an assessment, telling her that "Amira was not listening to her." Fleming, who initially thought her daughter was being bullied by an older student, said the software never produced reliable data for the entire first semester of first grade. She is part of a group of roughly 45 parents, mostly in Santa Fe and Albuquerque, who are opting their children out of the assessment.
Fleming said Vasquez's defense of the technology "came across to me as being much more uncritically enthusiastic about the use of Amira than I felt a representative from the PED should have been in light of all the criticisms." Committee members also raised questions about the tool's ability to understand students with accents or speech differences.
Privacy and data security questions
Privacy worries surfaced in Portland Public Schools last month when a board member noted that the contract included multiple checked-off boxes indicating the company would have access to student demographic data, including languages spoken, living situation, disability and migrant status. Amira's chief operating officer responded that the district had provided none of those metrics and that the data would only have been used for a research study requiring parental consent-a study that was never conducted. The company did not immediately explain why the boxes were checked.
At the Taos hearing, House Education Committee Chair G. Andrés Romero, an Albuquerque Democrat, asked Angel about data privacy. "We never, ever release or share our data with anybody other than the licensing educational agencies that we work with," Angel said. "We don't sell it." He described the company's "proprietary and private data center" and use of "advanced encryption," but gave no specifics on how student data from New Mexico is handled. He said the company would hold or delete data at the state's discretion.
Ifeoma Ozoma, a former policy lobbyist for Google and Facebook who now leads the Santa Fe parent effort, said she became an advocate for caution around AI after "seeing it from inside." She expressed doubt about the company's security capabilities and how the for-profit tech company uses student data to improve its models.
Policy gaps and procurement questions
The legislative committee's analyst, Saraí Ortiz, authored a report in late May noting that New Mexico lacks a statute governing AI vendor contracts, leaving "policy gaps" in how the state procures and deploys AI tools in schools. The report said it was unclear what vetting process the Public Education Department used before implementing Amira. The state began mandating the technology for K-2 students in the 2025-26 school year after Amira merged with Istation, the state's previous assessment provider, and the agency crafted a piggyback contract with the merged companies. Many districts also use the software for instruction.
Vasquez said the state's assessment contract would be open for bids in the 2027-28 school year, with other vendors able to apply to replace Amira. The controversy highlights the need for clear guidelines for AI for Education, especially as schools increasingly adopt automated tools. As lawmakers examine these issues, resources like an AI Learning Path for Policy Makers can help officials understand the technology's implications and craft effective oversight.
Competing claims on benefits and risk
The committee's report cited a January study from the Brookings Institution's Center for Universal Education that found the risks of artificial intelligence on children's development outweighed potential benefits. Angel pointed to Carnegie Mellon research from 2000-the basis for Amira's product-that reported automated speech recognition outperformed human teachers in certain areas and that students using the virtual program outperformed peers.
Rep. Joy Garratt, an Albuquerque Democrat and former teacher, questioned the comparison. "What I'm hearing from you, you feel like the Amira system, where the student is speaking into a mic, is more accurate in testing a child than a human teacher," she said, adding that she had heard from teachers who needed to spend extra time verifying the technology's findings. Angel replied that while an individual teacher might be more accurate in a given moment, "Amira [will] do the same thing for every one of the 50,000 students in New Mexico identically."
For parent Cassie Fleming, the trade-off is not reassuring. "You have to learn to read," she said. "And I'm just so disheartened that we're outsourcing that to an AI program - to a screen."
Why this matters for Education professionals
The New Mexico case shows that AI reading assessments, even when backed by state mandates, face sharp scrutiny over equity, data privacy, and the reliability of automated scoring for diverse learners. Educators and administrators should demand clear contracts that spell out exactly what student data is collected, how it is stored, and when it is deleted. They should also build in time for teacher verification of AI-generated results and push for independent evaluations of the tool's effect on reading outcomes, rather than relying on vendor-funded studies. The controversy offers a blueprint for the questions that should be asked before any district adopts voice-recognition AI for high-stakes assessments.
Your membership also unlocks:
