The Short Answer
More than 20,000 K-12 students joined the Presidential AI Challenge, but most school AI tools run on vendor clouds where student data leaves the district entirely. ibl.ai is the agentic AI platform where you own all the code and the data, self-hosted inside district infrastructure, model-agnostic across any LLM, and usage-based with no per-seat pricing — so student records stay inside the boundary by architecture.
COPPA and FERPA are not satisfied by a vendor's promise about how it handles data. They are satisfied by knowing where the data is.
A district that hosts its own AI platform answers that question with a rack diagram instead of a contract clause.
How big was the Presidential AI Challenge, and why does it matter?
More than 20,000 K-12 students participated in the Presidential AI Challenge, spanning all 50 states. It's an impressive number — a generation of students building AI literacy from early ages.
The initiative itself is exactly the kind of forward-looking investment education needs.
Its reach went further than the state count suggests: participation also covered the District of Columbia, Puerto Rico, and 49 Department of Defense schools across 10 countries.
The program grew out of an executive order signed in April 2025, with teams submitting projects in January 2026, state champions named in March, and finalists presenting at the national championship in Washington, D.C. in June 2026.
Students were asked to build AI solutions to real problems in their own communities, across elementary, middle school, high school and educator categories.
That framing is the right one. The concern is not that students are using AI — it is what happens to the data they generate while doing it.
Where does student data go when a school uses an AI tool?
There's a structural problem hiding under the headline: most school AI tools run on vendor clouds where student data leaves the district entirely.
Every query a student sends, every interaction logged, every performance metric recorded — it travels to servers the school district doesn't own or control. COPPA and FERPA aren't optional frameworks. They're federal law.
The breach data shows this is not a theoretical exposure. Clever's Cybersecure 2026 report found that 52% of U.S. school districts experienced a cybersecurity incident in 2025.
More telling is the shift in where those incidents originate. Vendor-related incidents rose from 4% of all K-12 breaches in 2023 to 32% in 2025 — an eightfold increase in two years.
The exposure also arrives unannounced. Shadow ed-tech — tools adopted by individual teachers without district IT review — is endemic in K-12, and it creates FERPA and COPPA obligations the district does not know it has.
What do COPPA and FERPA now require of school AI tools?
The compliance floor rose in 2026, and most AI procurement language has not caught up with it.
The amended COPPA Rule entered full enforcement on April 22, 2026. It requires separate verifiable parental consent for third-party disclosures, and it classifies biometric identifiers as personal information.
Consent can no longer be assumed or bundled. A district has to ask explicitly and document the decision, which is difficult when the data path runs through sub-processors nobody enumerated.
FERPA's School Official Exception permits a vendor to handle education records — but only under direct institutional control, for a legitimate educational interest, and without secondary use. An AI vendor that retains prompts for model improvement is testing that boundary.
Layered on top are more than 130 state student data privacy laws, several of which impose stricter deletion, notification and residency requirements than the federal baseline.
| Question a district must answer | Vendor-cloud AI tool | District-hosted platform |
|---|---|---|
| Where do student prompts land? | Vendor infrastructure | District infrastructure |
| Who are the sub-processors? | Disclosed by contract, may change | None for inference |
| Can records be deleted on request? | Via vendor process | Directly, in your database |
| What happens if the contract ends? | Access ends | System keeps running |
What does safety-first AI infrastructure look like in a district?
AI in K-12 needs safety-first infrastructure, not bolt-on compliance. That means on-premise or district-controlled deployments where student data stays within institutional boundaries by architecture, not by policy promise.
Encouraging a generation of students to engage with AI while routing their data through uncontrolled third-party systems isn't a calculated trade-off — it's an oversight that will eventually surface as a compliance crisis.
In practice, four properties define the safe version. Student prompts and outputs are stored in district-owned systems. Model inference runs inside the district's network boundary or an approved private environment.
Every AI interaction is logged where the district's own auditors can read it, not summarized in a vendor dashboard. And the platform outlives the procurement cycle, so a budget change never orphans a year of student work.
There is also a cost argument that district CFOs recognize immediately. Per-seat AI licensing bills on enrollment — a 20,000-student district pays for 20,000 seats whether or not a given student logs in that month.
Usage-based pricing tracks actual activity instead, which is the only shape that survives a district budget review at scale.
Can a district run AI on infrastructure it owns?
Yes — and for K-12, ownership is what turns compliance from an ongoing negotiation into a settled fact. With ibl.ai you own all the code and the data.
The full platform runs under a perpetual license on district or state infrastructure, so student records never cross an institutional boundary in the course of normal use.
It is model-agnostic, meaning a district can run an open-weight model entirely inside its own network, and change models later without changing platforms or renegotiating a data agreement.
Pricing is usage-based with no per-seat pricing. Deploy anywhere: your cloud, your VPC, on-premise, or a fully air-gapped network.
More than 1.6M users across 400+ organizations run the platform this way, including NVIDIA, MIT, and Syracuse University.
ibl.ai is family-owned and operated from New York, NY — a U.S.-headquartered, domestically-owned long-term partner, not a vendor that sells licenses and moves on.
20,000 students building with AI is a genuinely good outcome. Making sure the record of that work stays in the district that educated them is the part still left to do.