Parents, Not Policy, Paused the $57,590 Classroom Robot

Last updated: July 28, 2026

A small New York district spent $57,590 on a lifelike humanoid robot for its high school and planned to put it in front of roughly 500 students this fall. On July 24, the plan went on hold. What stopped it was not a state law or a district policy, because none exists for humanoid AI in classrooms. It was a news story, alarmed parents and teachers, and one letter from the state education commissioner. If an AI pilot is headed for your district next, that gap between adoption speed and actual guardrails is the part worth understanding.

TL;DR

Salamanca City Central School District in New York paused its $57,590 Realbotix humanoid robot pilot on July 24, 2026, hours after a letter from State Education Commissioner Betty Rosa.
The robot, nicknamed Sally, and a 24/7 AI tutoring avatar called Optio were scheduled to reach about 500 high school students in September 2026.
State officials cited unwritten privacy promises, unclear data retention, and inadequate parental consent procedures.
On July 27, 2026, New York State United Teachers asked the state for enforceable statewide guardrails on classroom AI, with heightened protections for humanoid systems.
No published research currently shows humanoid classroom robots improve K-12 learning outcomes.

A New York district paused its $57,590 humanoid classroom robot after parents, teachers, and the state stepped in. Here is what happened, and what to ask before an AI pilot reaches your child’s school.

Common questions

Did Salamanca cancel the classroom robot program?

No. The district paused the pilot on July 24, 2026 while it works with the New York State Education Department on enhanced student data privacy agreements and gathers more community input. Resumption stays on the table if state concerns are resolved. Realbotix stands behind the program and describes Salamanca as its flagship deployment, with more classroom robots planned.

Are AI tutors and classroom robots bad for kids?

The honest answer is that the evidence is thin in both directions. No published research shows humanoid classroom robots improve learning outcomes, and none proves broad harm either. Around-the-clock homework help and translation have genuine appeal, especially for families priced out of private tutoring. The distinction that matters: a tool that supplements human instruction is a different proposition from one that quietly substitutes for it, because children build skills through explicit teaching, feedback, and relationships with people who notice them.

What should I ask before my child’s school adopts AI technology?

Four questions cover most of it. Is there a written data agreement, not a verbal assurance, covering collection, retention, deletion, and access? What evidence of learning benefit exists, and what outcomes will the district measure? Is the tool a supplement to human instruction or a substitute for it? What is the consent process, and what happens for families who opt out? These are the same gaps New York officials cited in Salamanca.

What data do classroom AI tools collect about my child?

It varies widely by product, which is exactly why written agreements matter. In Salamanca, state officials flagged privacy promises that existed only verbally, unclear data retention, and uncertainty about who accesses student information and whether student interactions are used to improve the AI. Ask your district to produce the written data privacy agreement for any AI tool your child uses. A district that struggles to produce one has answered your question.

A NY district bought a $57,590 humanoid robot for its high school. No law slowed it down. Alarmed parents, teachers, and one letter from the state did. That gap is the real story. Tweet Now Follow Learning Success

A $57,590 robot, a letter, and a Friday pause

Salamanca City Central School District in western New York, where the high school sits on the Seneca Nation’s Allegany Territory and about 40 percent of students are Native American, approved a contract in June with Realbotix for “Sally,” a lifelike M-Series humanoid robot, paired with a 24/7 AI tutoring avatar called Optio. The robot was set to assist coding, robotics, and AI classes for 11th and 12th graders, with access expanding to about 500 high school students in September. Superintendent Mark Beehler defended the pilot as a teaching tool: “We truly believe in the importance of human-to-human interaction, so there is absolutely no way that we will allow devices like this or humanoids to replace teachers.”

Teachers saw it differently. “Placing a humanoid robot into the classroom to interact with our children, I believe, is a massive step in the wrong direction,” said Lacey Pihlblad, president of the Salamanca Teachers’ Association, whose union also flagged the software’s security and how student data would be gathered and stored, and asked the district to pause.

On Friday, July 24, hours after State Education Commissioner Betty Rosa sent the district a letter of concern, the district announced exactly that: “Our pilot project with Realbotix is currently on hold while we work through enhanced student data privacy agreements.” Among the gaps state officials cited: privacy promises that were verbal rather than written, unclear data retention, uncertainty about who accesses student information, and inadequate parental consent procedures.

Author Quote "

Students should not serve as test cases for machines built to befriend children.

What the coverage gets wrong

The loudest coverage treats this as a story about one company’s corporate history and a robot’s unsettling appearance. That framing lets every other district off the hook. The record that matters for parents is the adoption pattern: education systems have a documented habit of taking up practices without evidence and keeping them for decades. Learning styles were shown ineffective in 2008 (Pashler and colleagues) and again in 2024, yet a 2020 review across 18 countries found nearly 9 in 10 educators still teach to them. Guessing-based reading instruction persisted decades past Keith Rayner’s eye-movement evidence until a 2019 investigation forced new laws. The lesson of Salamanca is not that one robot was stopped. It is that parent and teacher scrutiny is, right now, the only guardrail between unproven technology and your child’s classroom.

Schools adopt first and check the evidence later. That pattern has a record.

Much of the coverage has focused on the robot’s unsettling looks and the vendor’s corporate history. The more useful frame for parents: nothing in state law required evidence of learning benefit before a lifelike AI reached a classroom, and nothing required written data protections either. That is not a Salamanca problem. It is how education technology adoption works, and the record shows what happens when it goes unchecked. Researchers established in 2008 that matching instruction to a child’s “learning style” does not improve learning, a finding independently reconfirmed in 2024. A 2020 review of educators across 18 countries found nearly 9 in 10 still teach to learning styles. Eye-movement research spanning decades showed that teaching children to guess words from pictures trains them to read like struggling readers, yet it took a 2019 investigative podcast to start changing reading laws. The system that kept those practices alive is the same system now buying humanoid robots on promises. When this pilot launched in June, we asked what the technology actually addresses in a brain that learns differently. The state is now asking its own version of that question.

The evidence cuts both ways, and honesty requires saying so. There is no published research showing humanoid classroom robots improve learning outcomes, and there is also no body of evidence proving harm. The vendor’s equity pitch names a real problem: private tutoring at $50 an hour is out of reach for many families, and a 24/7 homework helper has genuine appeal. An untested robot is not yet an answer to that problem. It is an experiment, and experiments on children carry a burden of proof that a press release does not meet.

That is what makes the teachers union’s response the sharpest reading of this story. New York State United Teachers wrote to state education leaders on July 27 asking for enforceable statewide guardrails on classroom AI, with heightened protections for systems built to resemble people or mimic human relationships. Union president Melinda Person put the gap plainly: “What stopped this robot was a news story, alarmed parents and educators, and a commissioner’s letter — not any law or policy of this state, because none exists.”

Key Takeaways:

No law did this: New York has no statute on humanoid AI in classrooms; parents, teachers, and one letter paused the pilot.

Adoption outruns evidence: Learning styles were debunked in 2008, yet a 2020 review found 9 in 10 educators worldwide still teach to them.

Ask before the pilot: Written data agreements, outcome evidence, and supplement-not-substitute status are the three questions for any district.

The playbook that worked, and the questions that travel

Here is the encouraging part. Nobody will ever advocate for your child as hard as you will, and in Salamanca that advocacy did what no policy could. Parents and teachers asked pointed questions in public, a reporter wrote them down, and a $57,590 deployment stopped to answer them. That playbook travels to any district, and it works before a school board vote even better than after one.

The questions worth bringing to any AI pilot are the ones the state ended up asking here. First, is there a written data agreement, not a verbal assurance, covering what is collected, how long it is kept, who has access, and whether student interactions are used to improve the vendor’s AI. Second, what evidence of learning benefit exists, and what outcomes will the district measure beyond engagement. Third, is the tool a supplement to human instruction or a quiet substitute for it. Fourth, what does the consent process look like, and what happens for families who opt out.

None of this argues for banning technology from classrooms. Students should learn how AI works, and tools that reduce homework friction have a place. The argument is about sequence: written safeguards and evidence first, deployment second. Realbotix describes Salamanca as its flagship deployment and has said more classroom robots are planned, so versions of this decision are coming to other school boards soon. The Salamanca pause shows the current guardrail is you.

Author Quote "

Many parents, teachers, and stakeholders are understandably concerned about the presence of artificial intelligence technology in classrooms, particularly when it involves robots that resemble humans.

The Salamanca pause proves a point worth keeping: nobody will ever advocate for your child as hard as you will, and this month that advocacy did what no state policy could. The villain here is not a robot or the people who bought it. It is a system that adopts unproven tools first and asks for evidence later, decade after decade. Your scrutiny is the guardrail, and it works best when you know what actually builds a child’s skills. Learning Success All Access gives you the science-backed programs and parent training to build those skills at home, no press release required.