The Salamanca City Central School District in upstate New York has paused a planned rollout of a nearly $60,000 humanoid robot named “Sally” after state education officials, teachers, and parents raised alarms about student privacy and the device’s classroom role. The Realbotix robot, intended for high school STEAM courses, is now on hold while the district renegotiates data-privacy terms with the New York State Education Department.

The decision marks a critical checkpoint for schools eyeing artificial intelligence in the classroom — and a vivid reminder that any AI endpoint, whether a robot or a Windows laptop, demands rigorous governance before it meets students.

Inside the Pause: What Triggered the Stop

Salamanca’s board approved the purchase earlier in 2026 with a straightforward pitch: “Sally” would help students learn Arduino programming, robot maintenance, and troubleshooting — not replace teachers. Superintendent Mark Beehler envisioned the robot as a motivator, exposing rural students to emerging technology without a two-hour drive to the nearest city.

But details from presentations blurred those lines. State Education Commissioner Betty Rosa noted in a letter that district materials described Sally as a “tutoring platform,” contradicting assurances that the robot would not deliver instruction. That single word transformed the conversation from a hardware demo into a debate over AI-driven teaching — and the data it might collect.

Teachers amplified the concern. The New York State United Teachers union zeroed in on Realbotix’s corporate connections to Intima LLC, which holds a stake in the sex-doll maker RealDoll. “A robot built by a company associated with sex dolls has no business in our classrooms,” union president Melinda Person said. Realbotix insists the subsidiaries operate independently, but the reputational cloud stuck.

By late July 2026, the district posted a statement on Facebook: the pilot would not proceed until “enhanced student data privacy agreements” were in place, and more community outreach was done. No one had pulled the plug permanently, but Sally would not enter a classroom for the 2026–27 school year as planned.

Why It Matters for Schools and IT Admins

For IT professionals managing Windows environments in education, the pause is more than a local uproar. It underscores the reality that every AI-powered device — whether it runs on a specialized OS or taps into Azure AI services — is an endpoint that must be governed.

The privacy puzzle. The district says Sally will not collect personal information, record audio or video, transmit data to Realbotix, or search the public internet. All content is stored locally. Realbotix adds that student authentication would use the district’s own systems and data remains encrypted. But as the windowsforum.com analysis noted, “vendor assurances are not a substitute for enforceable data-processing terms.” Schools need contracts that answer: exactly what data is generated during a student’s conversation with Sally? Where is it stored, for how long, and who can access it? Can the vendor use interaction logs for model training? What happens during a security incident?

Compatibility with existing infrastructure. Most U.S. schools run on Microsoft 365 with Azure Active Directory for identity, Intune for device management, and Defender for endpoint security. A humanoid robot doesn’t natively fit into that stack. IT teams would need to define how Sally joins the network (if at all), what compliance policies apply, and how to patch its software. If the robot’s content stays local but receives updates via USB or a restricted VLAN, those processes must be documented and auditable.

The trust factor. Even if technical controls are sound, public perception can derail a project. Teachers and parents in Salamanca voiced fears not just about privacy but about the manufacturer’s corporate history. For IT leaders, early transparency — demos for the school board, Q&A sessions with families, clear documentation — can prevent a similar backlash. A robot that looks disarming but comes from a controversial lineage may need extra effort to gain acceptance, no matter how locked down it is.

How We Got Here: The Road to This Moment

The Salamanca saga didn’t emerge in a vacuum. It reflects a broader, accelerating tension as AI enters schools.

Precedents and parallels. In 2025, a San Diego charter school rolled out two AI-powered robots from Engineered Arts at a cost of $500,000. Those machines, too, sparked debate about replacing human interaction. Globally, chatbots and AI tutors are being pitched as solutions to teacher shortages and personalized learning gaps. New York’s scrutiny of Sally builds on a patchwork of state-level student-data privacy laws, including Education Law §2-d, which restricts how vendors can handle personally identifiable information.

The Realbotix factor. The company, based in Las Vegas, has positioned itself for healthcare and education markets with its humanoid robots. The same subsidiary that built Sally also ships robots for companionship and retail. But its sister company’s ties to RealDoll, a longtime maker of hyper-realistic sex dolls, created an optics crisis. Realbotix stated that “no crossover” exists in products, management, or facilities. Superintendent Beehler, reached through a former colleague, defended the choice by noting Realbotix’s integration with WozEd, a curriculum from Apple co-founder Steve Wozniak that Salamanca already uses.

Union and community pushback. NYSUT framed the pilot as a step toward replacing human educators. Beehler countered that “teaching is a human-to-human process,” and Sally would only augment STEAM learning. But for some residents, like 2024 graduate Joplin Ficek, any AI presence felt intrusive: “They need to hire more people who are willing to put in the effort for the children.”

What to Do Now: An Action Plan for Schools Considering AI Devices

If your district is eyeing a Sally-like deployment — or any AI-driven classroom tool — here’s how to avoid a similar pause.

  1. Start with a privacy impact assessment. Map what data the device could generate: voice prints (even if not recorded), text logs, interaction timestamps, inferred student proficiency levels. Classify each element and identify legal obligations under FERPA, COPPA, and state laws.

  2. Demand a data processing agreement (DPA) that goes beyond promises. The contract must spell out:
    - Exact data categories collected or processed.
    - Storage location and retention periods.
    - Prohibition on secondary use, including model training.
    - Vendor access controls and encryption standards.
    - Incident notification timelines (e.g., 24 hours).
    - Audit rights, so you can verify compliance.

  3. Integrate the device into your existing endpoint management strategy. Even if Sally runs a custom OS, treat it like any other managed device. Determine whether it can be domain-joined or if it will live on an isolated VLAN. Use Microsoft Intune or a third-party MDM to enforce compliance policies if supported; otherwise, document manual update and patch procedures.

  4. Test community appetite early. Before a board vote, host demonstrations, invite questions, and publish a plain-language FAQ. Address reputational concerns about vendors head-on — don’t wait for the media to unearth them.

  5. Define the instructional role in writing. Clarify whether the device supplements, tutors, or merely demonstrates. That definition governs privacy rules: a tool that “tutors” may create protected education records; a static demo may not. Train teachers on what the device should and shouldn’t be asked, especially around sensitive topics.

  6. Plan for the end-of-life. What happens to student data when the robot is decommissioned or the contract ends? Specify data destruction methods and a deadline.

Looking Ahead: What’s Next for Sally and AI in Education

Salamanca’s pilot is paused, not cancelled. If the district can negotiate contractual safeguards that satisfy the state and community, Sally could still find a spot in a high school lab. The New York State Education Department’s engagement suggests regulators are watching — and may set a template other districts will follow.

For Microsoft-powered schools, the episode reinforces that AI governance is not just a cloud problem. Any device that interacts with students — whether a Windows laptop running Copilot or a humanoid robot — extends the data boundary. The hard work isn’t in the hardware; it’s in the policies, contracts, and trust-building that must precede it.